05 - radiol clin n am 2007 - update on colorectal cancer imaging

R A D I O L O G I CC L I N I C S

O F N O R T H A M E R I C A

Radiol Clin N Am 45 (2007) 85–118

85

Update on Colorectal CancerImagingMarc J. Gollub, MD*, Lawrence H. Schwartz, MD,Tim Akhurst, MD

- ScreeningFecal occult blood testSigmoidoscopyFecal DNA analysisBarium enema: time to put it to rest?ColonoscopyCT colonography (virtual colonoscopy)

- Diagnosis and stagingBarium enema

CTVirtual colonoscopyNewer imaging methods

- Primary and recurrent rectal cancerPositron emission tomography in rectal

cancer- Radiologic follow-up and treatment

monitoring- Summary- References

Worldwide, colorectal cancer (CRC) is the thirdmost frequently occurring cancer in both sexes,but it ranks second in developed countries [1]. Inthe United States, cancer is the second most com-mon cause of death after heart disease and causesone in four deaths. The American Cancer Society(ACS) estimates 1,399,790 new cases of cancer in2006. About 148,610 of these will be of colon or rec-tum. An estimated 55,170 deaths caused by CRC areexpected in 2006. Because of screening and prema-lignant polyp removal, CRC incidence rates havebeen decreasing since 1985. Because of improve-ments in survival, mortality rates have also beendecreasing an average of 1.8% per year. If CRC is di-agnosed at an early stage, the prognosis is favorablewith 5-year survival rates exceeding 90% [2].

The risk of CRC increases with age, with mostcases diagnosed after age 50, and a median age inthe mid-70s. The lifetime risk for CRC is approxi-mately 5% to 6%. Several risks factors for the devel-opment of CRC have been identified that could be

0033-8389/07/$ – see front matter ª 2006 Elsevier Inc. All rigradiologic.theclinics.com

altered, including obesity, physical inactivity, smok-ing, heavy alcohol consumption, a diet high in redmeat, and inadequate intake of fruits and vegetables.Protective effects may be gained from regular use ofnonsteroidal anti-inflammatory drugs, includingaspirin, estrogen and progestin hormone therapy,and 3-hydroxy-3-methylglutary coenzyme A reduc-tase inhibitors (cholesterol-lowering drugs). Thesedrugs, however, are not currently recommendedfor prevention [3].

More than 80% of CRC cases arise from adeno-matous polyps; however, less than 1% of adenoma-tous polyps smaller than 1 cm become cancer.Polyps 10 to 20 mm in size have an approximately4% risk of carcinoma and a 21% risk of high-gradedysplasia. Subcentimeter polyps have a much lowerrisk (%1%) for carcinoma and high-grade dysplasia(3%–5%) [4].

About 80% of CRC occurs in patients at averagerisk (no known risk factors and age 50 or greater)and 20% occur in those with a family history of

Department of Radiology, Weill Medical College of Cornell University, Memorial Sloan-Kettering CancerCenter, Room C276F, 1275 York Avenue, New York, NY 10021, USA* Corresponding author.E-mail address: [email protected] (M.J. Gollub).

hts reserved. doi:10.1016/j.rcl.2006.10.003

mailto:[email protected]

Gollub et al86

CRC in a first-degree relative. Of this latter group,a small proportion (6%) is associated with geneticsyndromes, such as familial adenomatous poly-posis and hereditary nonpolyposis CRC. Others athigher risk include those with long-standing ulcer-ative colitis and those with a personal history oflarge adenomatous polyps or CRC or a family his-tory of adenomatous polyps diagnosed before age60.

There have been exciting new developments inCRC research since the last monograph on CRCin the Radiologic Clinics of North America [5]. Mostare beyond the scope of this article:

1. Further understanding and delineation of mo-lecular and genetic details in the adenoma-carci-noma sequence of mutations, includingmicrosatellite instability markers and DNA mis-match repair genes

2. Development of fecal DNA assays for screening3. Development of more effective chemotherapy

[6]4. Rapid development and maturation of CT colo-

nography (CTC, also known as virtualcolonoscopy)

5. Rapid technologic evolution of multislice helicalCT imaging and higher field strength MR imag-ing magnets (3 T) with phased array coils

6. New and improved methods of liver metastasisdiagnosis and treatment, such as microbubblecontrast agents in ultrasound and radiofre-quency ablation techniques in interventional ra-diology, respectively

7. Exponential increase in understanding, use, andreimbursement for positron emission tomogra-phy (PET), now combined with CT (fusiontechnology)

8. Greater publicity and public awareness of thispreventable cancer killer through more wide-spread screening efforts and media coverage.

Screening

Screening persons at risk for CRC to detect precan-cerous polyps differs from screening for tumors likebreast or cervical cancer. Whereas the mammogramand the Papanicolau smear are single examinationsthat are widely accepted and used and for whichlittle overall controversy remains, for CRC therecontinues to be several available screeningexaminations and various different strategies rec-ommended by several health care advisory commit-tees (Table 1).

Existing clinical trials for CRC screening have notdirectly compared different screening approachesfor clinical effectiveness or cost-effectiveness andhave not tested starting and stopping ages. In an

analysis of seven publications using simulationanalyses, Pignone and coworkers [7] found thatany of the commonly recommended screening strat-egies for adults age 50 years or greater reduce mortal-ity from CRC. The cost per life-year saved for CRCscreening ranges from $10,000 to $25,000 and com-pares favorably with other commonly endorsedpreventative health care interventions, such asscreening mammography. There have been no con-clusions on the most cost-effective strategy.

Various options exist because no single test showsunequivocal superiority. Although this allows pa-tients a choice, perhaps promoting greater compli-ance, these choices might result in confusion. Thismight in part explain why less than 40% [8] ofscreen-eligible patients in the United States haveever undergone CRC screening. More probable ex-planations for underscreening include embarrass-ment, fear, and aversion to the colonic cleansingrequired for some examinations [9].

Recent increases in public awareness, in partthrough the publicity surrounding sports personal-ities like Darryl Strawberry and media personalitieslike Katie Couric, have finally brought screening forCRC to the forefront. Nonetheless, efforts must stillincrease exponentially to help recruit the millionsof people at risk who still go unscreened.

Indirect studies have demonstrated that as manyas 95% of cancers arise from colorectal adenomasalong the adenoma-carcinoma sequence [10]. Be-cause this process has been estimated to take an av-erage of 10 years to occur (polyp dwell time), thereis ample time during which screening can be per-formed, and ample opportunity to prevent deathscaused by this leading cancer killer [11].

For the estimated 12,000,000 American adultseach year [12] who become eligible for screening,by virtue of turning age 50, without other knownrisk factors (so-called ‘‘average risk’’), numerous op-tions are available, including the fecal occult bloodtest, sigmoidoscopy, colonoscopy, barium enema,fecal DNA assay, and CTC.

Fecal occult blood test

Four randomized controlled trials of serial fecal oc-cult blood test conducted in Minnesota, the UnitedKingdom, Sweden, and Denmark involving morethan 300,000 subjects followed for up to 18 yearshas consistently demonstrated that serial fecal oc-cult blood test reduces colorectal mortality from15% to 33% [10,13]. The test suffers from manyfalse-positive and false-negative results, however,with a reported sensitivity for detection of CRC inthe range of 27% to 57%, and as low as 8% for ad-enoma [14]. Rare red meat and some vegetablesand fruits containing peroxidase can cause false-positive results. Nonbleeding CRC and polyps

Update on Colorectal Cancer Imaging 87

Table 1: Screening recommendations for colorectal cancer and polyps

Risk category Screening method Age to begin screening

Average risk Choose one of the following 50 yearsFecal occult blood testing annuallyFlexible sigmoidoscopy every 5 yearsFecal occult blood testing annuallyand flexible sigmoidoscopy every 5yearsa

Double-contrast barium enemaevery 5 to 10 yearsb

Colonoscopy every 10 yearsFamily history Choose one of the following 40 years or 10 years before cancer

was diagnosed in the youngestaffected family member,whichever is earlier 5 years

Colonoscopy every 10 yearsDouble-contrast barium enemaevery 5 to 10 years

Hereditarynonpolyposiscolorectal cancer

Colonoscopy every 1 to 3 years 21 years

Genetic counselingConsider genetic testing

Familialadenomatouspolyposis

Flexible sigmoidoscopy orcolonoscopy every 1 to 2 years

Puberty

Genetic counselingConsider genetic testing

Ulcerative colitis Colonoscopy with biopsies fordysplasia every 1 to 2 years

Seven to eight years after thediagnosis of pancolitis12 to 15 years after thediagnosis of left-sided colitis

a Some experts recommend combining annual fecal occult blood testing with flexible sigmoidoscopy every 5 years.b Rigid proctoscopy is recommended as an adjunctive examination to allow adequate visualization of the distal rectum.Flexible sigmoidoscopy may be necessary to evaluate tortuous or spostic signoid color.From Read TE, Kodner IJ. Colorectal cancer: risk factors and recommendations for early detection. Am Fam Physician1999;59:3803–92; with permission. Copyright ª1999 American Academy of Family Physicians. All rights reserved.

(the majority) and tumors in persons ingesting vita-min C can cause false-negative fecal occult bloodtest results. Most people with a positive test donot have CRC, but are subjected to the risk, cost,and discomfort associated with colonoscopy. Fur-thermore, compliance rates in the largest trialswere below 60% [11,14]. Recent evidence, compar-ing the immunohistochemical tests for a singlemonoclonal antibody agglutinating with hemoglo-bin A showed a higher specificity and sensitivity forhuman hemoglobin than the guaiac-based fecal oc-cult blood test [14]. Finally, stool obtained duringa digital rectal examination is inadequate to screenfor fecal occult blood, and despite its widespreaduse, this type of testing is not recommended.

Sigmoidoscopy

Flexible sigmoidoscopy provides direct visualiza-tion of a portion of the colon and suspicious lesionscan be removed. It is a very safe procedure with only

two perforations reported in a retrospective reviewof 49,501 examinations [10]. Only 65% to 75% ofadenomatous polyps and 40% to 65% of CRC,however, are within the reach of the 60-cm flexiblesigmoidoscope. In fact, recent observational studiesfrom colonoscopy suggest that one half of all ad-vanced adenomas (>1 cm) and cancers in the prox-imal colon are missed on sigmoidoscopy [10]. Thisexplains why this type of examination has been re-ferred to in the radiology community as tanta-mount to performing a unilateral mammogram toscreen for breast cancer [15]. The sensitivity of sig-moidoscopy for CRC and large polyps is 96% andfor small polyps 73%. The specificity for CRC andlarge polyps is 94% and for small polyps 92%.Nonetheless, no completed, large, randomized con-trolled trials have demonstrated the effectiveness ofsigmoidoscopy in the prevention of CRC death.Only indirect evidence from case-control studiessupports its effectiveness. Mortality reductions of

Gollub et al88

between 59% and 80% have been reported [13].Two ongoing studies, the National Cancer InstituteProstate, Lung, Colorectal and Ovarian screeningtrial and the UK Flexiscope Trial, will assess 60-cmflexible sigmoidoscopy in more than 250,000 sub-jects. The US Agency for Health Care Policy and Re-search recommends flexible sigmoidoscopy every 5years. Colonoscopy should follow sigmoidoscopywhen factors associated with an increased risk forproximal neoplasia are present, including age >65years, a polyp with villous histology, adenomaR1 cm, multiple distal adenomas, and a family his-tory of CRC. Current evidence suggests that the riskof advanced neoplasia in persons with only a distalhyperplastic polyp is not greater than those withoutdistal polyps.

The reduction in CRC mortality using a combinedapproach of fecal occult blood test and sigmoidos-copy has never been studied in a randomized trial,but two randomized controlled trials have reportedthat the addition of a one-time fecal occult bloodtest to sigmoidoscopy increased the detection ofCRC from 70% to 76%. The ACS and other agenciesrecommend a yearly fecal occult blood test in com-bination with sigmoidoscopy every 5 years as onemethod of screening. This option continues to beviewed as inferior if colonoscopy resources areavailable, because direct visualization of the entirecolon is only possible with colonoscopy [11].

Fecal DNA analysis

The molecular genetics of CRC were initially popu-larized by molecular biologists in the late 1980s[16]. Eighty-five percent of CRC result from chro-mosomal instability, with mutations involving theadenomatous polyposis coli gene, the p53 tumor-suppressor gene, and the K-ras oncogene. The other15% arise from loss of genes involved in DNA mis-match repair, manifested by microsatellite instabil-ity. A recent study comparing a fecal-based,multitarget DNA panel with Hemoccult II in asymp-tomatic adults 50 years of age or older who were ataverage risk for CRC revealed a sensitivity four timesthat of the Hemoccult test (51.6% versus 12.9%, P 5.003) for invasive cancer and more than twice assensitive as the Hemoccult for advanced adenoma(40.8% versus 14.1%, P < .001) without a loss ofspecificity (94.4% for fecal DNA versus 95.2% forHemoccult II) [8]. The increased detection of non-advanced adenoma from 11% to 15% was not sig-nificant. In previous studies, not exclusivelyconfined to asymptomatic patients, the sensitivityfor detection of cancer has ranged from 37% to71%. In this and other studies, most lesions detectedby colonoscopy were not detected by either test. Al-though colonoscopy is superior to other tests, thePreventive Services Task Force has determined that

no single test or strategy for CRC screening can beendorsed on the basis of currently available data. Al-though the fecal DNA test has a low sensitivity, po-tentially limiting its use as a one-time test for cancer,like the Hemoccult test, its use at more frequent in-tervals might be as effective as a more sensitive testused infrequently, such as colonoscopy. More stud-ies are needed at this time. The ACS, at the time ofthis writing, does not yet include this test in itsscreening recommendations. Reasons may include(1) a wide confidence interval of sensitivity of35% to 68% at a 95% probability; (2) higher cost;and (3) acceptability (patients must collect and re-frigerate an entire bowel movement).

Barium enema: time to put it to rest?

Double-contrast barium enema (DCBE) is includedin the screening recommendations of the ACS at5-year intervals. The rationale for its inclusion in-cludes its cost-effectiveness [17] and wide availabil-ity. It is less sensitive at detecting colonic neoplasmsthan colonoscopy, and cannot remove polyps. Al-though to date randomized controlled trials arenot available to determine if screening by DCBEreduces mortality from CRC, DCBE detects mostadvanced adenomas and cancers [10]. In theNational Polyp Study, however, the sensitivity forpolyps even as large as 1 cm or greater was approx-imately 50%. Many false-negative results are foundfor smaller polyps, but their clinical importance isquestionable [18]. A case-control study did reveala 33% reduction in CRC death, but confidence in-tervals were wide [19]. In a nonrandomized studyof 2193 consecutive CRC cases in a communitypractice the sensitivity of DCBE for cancer was85% compared with colonoscopy of 95% [20]. Fi-nally, an investigation that received a lot publicityand caused a lot of uproar in the radiology commu-nity [21,22], and which may have done more toharm the reputation of the DCBE than warranted,was a prospective study in a surveillance populationcompared with colonoscopy as a gold standard.Here, DCBE detected only 39% of polyps identifiedat colonoscopy, including 21% of polyps 5 mm orsmaller, 53% of adenomatous polyps 6 to 10 mmin size, and only 48% of those >1 cm in size.Some of the limitations of this study included thepresence of only 23 lesions greater than 1 cm in973 patients, the risk of extrapolating data froma surveillance population to a screening popula-tion, and the lack of generalizability of results de-rived from expert academic centers to communitypractice. Additionally, the high completion rate ofcolonoscopy in this study and the 3:1 ratio ofmale to female patients may not reflect a truescreening situation in the average community prac-tice. Use of sensitivity statistics from a one-time


screening examination also ignores the potentialfor subsequent detection of early lesions on repeatscreening studies [17]. Finally, as the authors ofthis same study admit, ‘‘screening and diagnostictests are judged not only on the basis of their accu-racy in detecting and ruling out lesions but also onthe basis of their safety, convenience, acceptance bypatients, cost and cost-effectiveness, as well as onthe number of physicians needed to conduct the ex-amination properly’’ [23].

A survey performed in collaboration with the Na-tional Cancer Institute, the Centers for DiseaseControl and Prevention, and the Centers for Medi-care and Medicaid Services published in 2002found evidence to support the claims of a continu-ing trend of colonoscopy replacing DCBE as an ini-tial colorectal examination since 1980. Althoughthe benefits of colonoscopy are well known andpublicized, in this survey of 1718 primary care phy-sicians and 381 nationally representative radiolo-gists, 75% of radiologists but only 33% ofprimary care physicians rated DCBE as a ‘‘very effec-tive’’ modality. Furthermore, despite high ratings byboth groups, fewer than 10% of radiologists andfewer than 10% of primary care physicians reportedthat colonoscopy should be or was the screeningapproach most often recommended to their pa-tients. Single-contrast barium enema is less sensi-tive than DCBE, and alone is not recommendedas a screening strategy [18].

Colonoscopy

There are no randomized control studies evaluatingwhether colonoscopy screening alone reduces theincidence or mortality from CRC in people at aver-age risk. But two large cohort studies, the ItalianMulticenter Study [24] and the National PolypStudy [25], revealed a reduction in CRC incidencecompared with nonconcurrent control groups.The National Cancer Institute is now sponsoringa study of colonoscopy screening (Prostate, Lung,Colorectal and Ovarian). The trial began in 1993and closed to accrual in 2001, with screening con-tinuing until 2006 and follow-up anticipated for10 years thereafter. The study includes nearly155,000 men and women aged 55 to 74 [26]. TheACS and others recommend screening colonoscopyevery 10 years based on data extrapolated from sig-moidoscopy studies showing a protective effect ofendoscopy up to 10 years [11]. In addition, Medi-care instituted reimbursement of screening colono-scopy as of July 2001.

The test is performed with a 160-cm flexible en-doscope. A recent large study of screening colono-scopy resulted in a completion rate of 98% ofpatients with a mean procedure time of 30 minutes[27], but completion rates of 75% to 96% have

been reported elsewhere [17,18]. The safety of thetest has been well-established, but the rate of per-foration and death is not trivial. Perforation orhemorrhage is reported to occur in 1 of 500 exam-inations with fatality in 1 of 5000 [28]. By compar-ison, for barium enema, important complicationsof any type arise in 1 of 10,000 cases and the perfo-ration rate is reported to be 1 of 25,000 cases withdeath in 1 of 55,000 [29]. Although colonoscopyis considered the criterion standard for detectingCRC and adenomas, it is an imperfect examination.Up to 6% of advanced adenomas were missed in anoft-quoted series of back-to-back examinations per-formed by two expert examiners. Miss rates were13% for adenomas 6 to 9 mm and 27% for thosebelow 5 mm [30]. More recent data using the newlydeveloped CTC have challenged previously undis-puted claims of excellent accuracy in colonoscopy.Using a method known as ‘‘segmental unblinding,’’a second colonoscopic inspection is made in a seg-ment with a positive finding at CTC. With this newstandard, Pickhardt and coworkers revealed a colo-noscopic miss rate of 10% for polyps R10 mm[31,32].

Reported sensitivities for colonoscopy are 96.7%for cancer, 85% for large polyps, and 78.5% forsmall polyps, with a specificity of 98% for all lesions[33]. Although it is the only technique that offersscreening, diagnosis, and therapeutic managementof the entire colon in one procedure, it involvesgreater cost, risk, and inconvenience to the patientthan other screening tests, all important features tobe taken into consideration for any screening test.

Finally, two recent trends and two new develop-ments warrant mentioning. First, misuse of this ex-pensive resource, specifically, unwarranted repeatcolonoscopy or short interval surveillance for insig-nificant hyperplastic polyps or even mucosal tags,has been documented by the US National CancerInstitute survey regarding postpolypectomy surveil-lance [34]. In an era in which waiting lists for colo-noscopy exist in parts of the country this practicemay result in even further decreased rates of screen-ing [35]. Second, the recent introduction of intra-venous propofol, a deep sedative hypnoticmedication, administered by an anesthesiologist(and adding significantly to the cost of the proce-dure) has resulted in fewer incomplete colonoscop-ies in the authors’ institution and likely ona national basis. Two new exciting colonoscopic de-velopments beyond the scope of this article includethe use of endoscopic mucosal resection, wherebyat colonoscopy solution is injected submucosallybefore polypectomy, and chromoendoscopy, inwhich characterization of mucosal topography isused to predict histology. The reader is referred toSaitoh and coworkers [36].

Gollub et al90

CT colonography (virtual colonoscopy)

This new technology (described further later) is notas yet approved for use as a screening test. No ran-domized controlled studies are available investigat-ing the ability of CTC to reduce CRC morbidity.Until recently, no screening studies were availableto evaluate the efficacy of CTC, and the accuracyof the test had to be extrapolated from multi-ple studies using primarily surveillance-typepopulations.

The appeal of CTC in the screening setting derivesfrom the fact that the examination is a noninvasiveCT scan that uses no sedatives or contrast media,and could provide an attractive alternative formany patients who refuse to or cannot undergo co-lonoscopy. Because it is an imaging test only, andcannot remove polyps, its role is limited comparedwith colonoscopy.

The first and largest screening study performed byPickhardt and coworkers [37] at three United Statesmilitary hospitals revealed that CTC performed aswell as or better than optical colonoscopy ina same-day back-to-back correlative comparisonof 1233 asymptomatic subjects at average risk forCRC. Five experienced radiologists and 17 experi-enced colonoscopists were involved. Sensitivity foradenomatous polyps was 94% for CTC versus92% for optical colonoscopy at the 8-mm diameterthreshold and 96% for CTC versus 88% for opticalcolonoscopy at the 10-mm threshold size. The accu-racy of CTC for adenomatous polyps on a per-patient basis was 92% for 8 mm and 96% for 10mm. CTC depicted 54 (91.5%) of 59 advanced neo-plasms, whereas optical colonoscopy depicted 52(88.1%). The negative predictive value of CTC was99% for adenomas 8 mm or larger. The authorswere the first to use a primary three-dimensional in-terpretation method after fecal tagging with bariumand meglumine diatrizoate, followed by electronicsubtraction of labeled stool and fluid using a com-mercially available, US Food and Drug Administra-tion–approved computer. The technique ofsegmental unblinding (see later) allowed separatevalidation of both CTC and colonoscopy. These re-sults closely mimicked or improved on a large num-ber of preceding studies performed in surveillanceor mixed-type populations, using combined two-di-mensional and three-dimensional interpretations,without fecal tagging and electronic subtraction(Table 2) [2,38,39]. A coincident study, not per-formed in a screening population and using oldertechnology with less reader experience [40], led tomuch argument and deliberation in the literature,and served well to point out the necessity of atten-tion to detail (ie, slice thickness, polyp nomencla-ture, and reader experience) required to achieve

robust and reliable results, and to point out the ex-istence of a steep radiologist’s learning curve forCTC [41]. Nonetheless, as of this writing, no otherlarge study in screening subjects has been per-formed to validate the Pickhardt study.

Other screening trials now in progress includea large multicenter trial launched by the AmericanCollege of Radiology Imaging Network to testtwo-dimensional versus three-dimensional accu-racy in 2600 subjects. The Special Interest Groupin Gastrointestinal and Abdominal Radiology(United Kingdom) will compare CTC with bariumenema and colonoscopy in 4500 patients. The Ital-ian Multicenter Study on Accuracy of CT Colonog-raphy will enroll 3550 patients.

Because of the novelty of this technique, cost-ef-fectiveness has only been able to be investigated us-ing various statistical models [42–44]. Taking intoaccount compliance, the cost of care from missedpolyps, the avoidance of perforations and other fac-tors, and assuming optimal sensitivity and specific-ity for CTC, some investigators have determinedthat the cost of CTC has to be significantly lowerthan colonoscopy (up to 54% less) to be cost effec-tive [42].

Numerous studies have looked at patient comfortand preference for CTC versus colonoscopy [45]and found varying results. In a recent study testingpreference immediately after each test and 5 weekslater, CTC with bowel relaxants was preferred overcolonoscopy [46].

Although CTC is noninvasive, potentially seriousadverse events have been reported in 0.08% ofsymptomatic patients and perforations have oc-curred in 0.05% to 0.059% [47,48]. By comparison,the colonoscopy perforation rate reported in thesame and similar hospitals was 0.13%, suggestinga much lower risk with CTC [47].

Finally, use of a screening test that uses radiationmust address concerns regarding potential radia-tion-induced cancers. According to one investigator,the best estimate for the absolute lifetime cancerrisk using the typical scanner (prone and supinescans, for one CTC only) is about 0.14% in a 50year old and half of that in a 70 year old. Multipleinterval examinations increase these values [49].

Diagnosis and staging

The diagnosis of CRC is definitively made at histo-pathologic examination using biopsy or surgicalspecimens, although radiographic appearances canbe pathognomonic. Most often, colonic malignancyis discovered through asymptomatic screening ex-aminations; during endoscopic examinations donefor symptoms; as incidental discovery on imagingtests; by clinical examination revealing signs of intes-


tinal blood loss (hematochezia, or tarry stools); or asa result of abnormal blood tests indicating anemiaor elevated carcinoembryonic antigen (CEA). Oncediscovered or suspected, confirmation and simulta-neous staging are the roles most often conferred onCT, MR imaging, ultrasound, CTC, and DCBE. Theradiologist must describe tumor size; location;depth of local wall penetration, if feasible; involve-ment of nodes; spread to other organs; and asso-ciated complications, such as obstruction,hemorrhage, abscess formation, inflammation, per-foration and the like (within the limitations of themodality). In the following subsections, each mo-dality is reviewed with respect to expected appear-ances and pitfalls in detecting and staging CRCand its relative accuracy in staging.

Staging of CRC has evolved to the use of tumor,node, metastasis system (Table 3), rather than theDuke’s classification or the modified Duke’s (As-tler-Collin’s) system. These latter systems are notdiscussed. The tumor, node, metastasis system re-sults in stage groupings for which prognostic andsurvival statistics are published (Table 4). The stagesand associated survival statistics are based on path-ologic staging not radiologic assessments (eg, pT1;pathologic T1 not uT1; ultrasound stage T1).

Barium enema

The role of the single-contrast barium enema or theDCBE in diagnosing and staging known or sus-pected CRC is threefold: (1) it can be used to diag-nose colon cancer in those patients undergoingscreening or work-up of symptoms; (2) it can beused to complete the total colon examination incases of incomplete colonoscopy, a role tradition-ally reserved for barium enema, but now largely re-placed by CTC; and (3) it can simply confirma suspected diagnosis of CRC from other imagingor clinical tests when colonoscopy is not available,not affordable, or not preferred by the patient. Al-though it is inferior to colonoscopy in the detectionof small polyps, large polyps, and cancers, it hasmany advantages including lower cost, lack of ne-cessity for sedation, greater safety, and wider avail-ability. Furthermore, because colonoscopy canoccasionally miss lesions, these two examinationsmay be viewed as complementary, the developmentof CTC notwithstanding. It is unfortunate that theexplosive technologic imaging revolution has re-sulted in a diminishing cadre of highly skilled radi-ologists who perform and teach these worthwhileexaminations. Nonetheless, with scrupulous atten-tion to good colonic cleansing and with rigorousfluoroscopic-radiographic technique, described indetail elsewhere [50], most polyps and cancerscan be detected with DCBE.

The detection of adenomatous polyps, whichmay undergo malignant transformation throughthe adenoma-carcinoma sequence, and the detec-tion of neoplasms is the primary goal. Most polypsnever become CRC [51], and the risk is relatedto polyp size. Only about 1% of polyps less than10 mm harbor CRC, whereas 10% to 20% of adeno-mas 10 to 20 mm in diameter and 40% to 50% ofthose greater than 20 mm in diameter harboradenocarcinoma.

Small, classic, hyperplastic polyps, often effacedby distention, are not important to detect, becausethey are not believed to undergo this transforma-tion (a recently described hyperplastic polyposissyndrome of ‘‘serrated’’ adenomas is one exception,but these are larger lesions) [52].

Adenomas are classified as tubular, tubulovil-lous, or villous at histologic examination. Thegreater the amount of the villous component, thegreater is the risk of malignant degeneration.Polyps may be flat, sessile, or pedunculated(Fig. 1). A pedunculated polyp with a stalk longerthan 2 cm is rarely associated with invasive carci-noma. On DCBE, sessile polyps on the dependentsurface appear as filling defects in a pool of bar-ium. Those on the nondependent surface areetched in white and appear as ring shadows(Fig. 2). These can at times be confused with diver-ticula, but the ‘‘bowler-hat’’ appearance with thedome of the hat pointed inward toward the colonlumen distinguishes these from diverticula [53].The most common entity confused with polyps isstool. Even in a well-prepared colon, one or twofragments of fecal residue may persist, requiringrotation and repositioning or gentle palpation ofthe patient (Fig. 3). Villous adenomas have a high-er risk of malignant degeneration. They are recog-nized by their granular or reticular appearancebecause of the filling of the interstices on the polypsurface (Fig. 4). Carpet-lesions, a particularly wor-risome diffuse type of villous tumor, may be hardto recognize and may be confused for stool. Thesemay protrude very little into the colon lumen [53].Flat-lesions are a diagnosis de rigueur of late, be-cause of the recognition that these may easily bemissed at colonoscopy and CTC. These are impor-tant and controversial lesions. It is estimated that36% of adenomas in European populations and22.7% of adenomas in the United States popula-tion may be flat [54,55]. Controversy surrounds re-ports that these are more likely to have advanceddysplasia compared with protuberant polyps.One problem, learned from the CTC literature, isthe inconsistent distinction of these from, for ex-ample, sessile lesions. A small study of flat adeno-mas found that their detection was limited atDCBE compared with colonoscopy [56].

Gollub et al92

Table 2: Literature survey of CT colonoscopy

Per polyp sensitivity %

N patients OverallDetection ofpolyps <6 mm

Detection ofpolyps 6–9 mm

Detection ofpolyps >9 mm

Rockey et al, 2005 614 — — 60 64Chung et al, 2005 51 90 84 94 100Cotton et al, 2004 600 12.7 7.6 22.7 51.9Macari et al, 2004 186 27.7 14.7 46.2 90.9Van Gelder et al, 2004 249 51.8 40.6 76.7 77.8Macari et al, 2004 68 21.4 11.5 52.9 100Hoppe et al, 2004 92 42.6 25.4 57.9 70.6Pickhardt et al, 2003 1233 — — 83.6 92.2Iannaccone et al, 2003 158 70.3 51.4 83.3 100Johnson et al, 2003 703 — — 47.1 46.3Pineau et al, 2003 205 46.8 29.4 75 77.8Taylor et al, 2003 54 48.4 37.5 75 100Ginnerup Pedersenet al, 2003

144 — — 73.7 92.3

Yee et al, 2003 182 69.9 60.3 79.8 92.7Munikrishnan et al, 2003 61 75.8 53.3 83.3 100Laghi et al, 2002 165 78.4 50 82.4 91.7Gluecker et al, 2005 50 22.4 2.4 33.3 81.8Lefere et al, 2002 100 77.5 56.5 90.3 100Macari et al, 2002 105 32.6 12.1 70.4 92.9McFarland et al, 2002 70 — — 36.1 68.1Yee et al, 2001 300 77.5 66.9 81.8 94.1Hara et al, 2001 237 — — — —Spinzi et al, 2001 96 57.8 — — 61.5Fletcher et al, 2000 180 60.1 — 47.2 75.2Morrin et al, 2000 81 — 32.9 64.5 90.9Mendelson et al, 2000 53 27.5 17.5 22.2 72.7Macari et al, 2000 42 37.5 20 60 100Morrin et al, 2000 34 — — — —Fenlon et al, 1999 100 71.3 66.7 89.7 90.9Rex et al, 1999 46 22 11.1 42.9 50Dachman et al, 1998 44 46.7 7.7 33.3 83.3Royster et al, 1997 20 91.4 66.7 90 100Hara et al, 1997 70 37.4 25.9 57.1 70

From Mulhall BP, Veerpan GR, Jackson JL. Meta-analysis: computed tomographic colonography. Ann Intern Med2005;142:635–650; with permission.

Most CRC are detected on DCBE, and are eithersemiannular or annular (Fig. 5). Fewer than 10%appear as polypoid or carpet lesions [53]. False-negative examinations do occur when lesions areoverlooked because of perceptive error fromoverlapping bowel loops (Fig. 6). Overlooked carci-nomas in the rectum may occur if the rectal tube orballoon obscures the mucosa. Attention to tech-nique avoids these errors (Fig. 7).

Rarely, other abnormalities may mimic annularCRC, such as amoeboma or histoplasmoma [57].In the authors’ oncologic population, slow-grow-ing secondary malignancies, such as ovarian can-cer, have occasionally been seen to mimic theannular appearance of primary CRC (Fig. 8). Be-cause 5% of patients with CRC have a synchronous

lesion, every attempt should be made to examinethe entire colon at DCBE when one nonobstruc-tive lesion is found. Single-contrast barium enemahas been shown to be inferior to DCBE. Althougha large annular lesion may be very well appreci-ated, smaller malignancies can be missed in thebarium pool even with systematic compression[58]. The use of single-contrast barium enema(or water soluble–based contrast enema) is best re-served for the elderly and infirm and those cases inwhich there is a question of obstruction, perfora-tion, or anastomotic leak. Barium enema playsa minimal role in the diagnosis of recurrentCRC. Most cases arise extraluminally next to ordistant from the regions of resection of the pri-mary tumor, rather than at the anastomosis. For


Per patient sensitivity % Detection of polyps 6–9 mm Detection of polyps of >9 mm

OverallPatientswith cancer

Overallspecificity % Sensitivity % Specificity % Sensitivity % Specificity %

— 78 — 51 — 59 96— — — — — — —20.5 75 90.5 30.3 93.1 54.8 95.9— — 83.1 — — — —62.1 — 30.6 — — 83.9 92— — 89.7 — — — 98.573.3 87.5 — — — 95 98.5— — — — — 93.8 9696 100 96.5 — — — —— — — 52.2 91.2 47.9 97.561.8 — 70.7 — — 90 94.664.5 83.3 — 50 — 90 100— — — 82.4 — 95.7 —

90.4 — 82.4 — — — —74.3 96.6 96.2 — — — —93 100 — — — — —— — 90.5 — — — —86 — — 91.3 92.2 100 10057.6 100 87 — — — —— — — 71.3 — 87.5 —93.9 100 56.5 95.2 — 100 —— — — — — 67.9 95.5— 87.5 — — — — —— — — — — 85.4 92.9— — — 72.7 96.6 87.5 100— — — — — — —— — — — — — —— 100 — — — — —82.4 100 83.7 94 92 96 9645.5 — — 42.9 — 80 88.943.8 — 89.3 — — — —— 95 — — — — —— — — — — 75 90.5

Table 2: (continued)

this reason, CT or PET is of greater use than lumi-nal investigations, such as barium enema orcolonoscopy.

CT

Extrahepatic diseaseCT technology has evolved rapidly from incremen-tal to single-slice helical to 4-, 8-, 16-, 64-, and 256-slice multidetector scanners now commerciallyavailable. Images obtained from 16-slice (orgreater) scanners can be near-isotropic (ie, nearlyperfectly cubical voxels without distortion of anat-omy). This allows for so-called ‘‘volumetric imag-ing’’ during a single breathhold in many instancescovering the entire chest, abdomen, and pelvis.The acquisition of ultrathin slices (typically 0.5,0.625, or 0.75 mm depending on the vendor)

allows reformation of images in any plane withoutloss of resolution (isotropic) [59]. This technologicadvance, allowing multiplanar postprocessing, putsCT on par with or ahead of (because of better spatialresolution) MR imaging. Nonetheless, a remainingadvantage of MR imaging is the lack of ionizingradiation.

Faster CT scanning requires faster injection rates,smaller volumes, the use of saline bolus chasers,more concentrated iodine contrast agents, and lon-ger delays between injection time and scan time.For a complete discussion on this topic the readeris referred to the review by Brink [60]. Various CTangiographic techniques can be used for preopera-tive liver assessment (discussed later).

Although CT scan is widely used for preoperativestaging of CRC, there is no consensus on its use in

Gollub et al94

Table 3: TNM staging system for colorectal cancer

Stage Definition

Primary tumor (T)TX Primary tumor cannot be assessedT0 No evidence of primary tumorTis Carcinoma in situ: intraepithelial or invasion of lamina propriaT1 Tumor invades submucosaT2 Tumor invades muscularis propriaT3 Tumor invades through muscularis propria into the subserosa or into

nonperitonealized pericolic or perirectal tissuesT4 Tumor perforates visceral peritoneum or directly invades other organs

or structuresRegional lymph nodes (N)a

NX Regional lymph nodes could not be assessedN0 No regional lymph node metastasesN1 Metastases in one to three regional lymph nodesN2 Metastases in four or more regional lymph nodes

Distant metastases (M)MX Distant metastases could not be assessedM0 No distant metastasesM1 Distant metastases

Extent of resection (R)b

RX Presence of residual tumor cannot be assessedR0 No residual tumorR1 Microscopic residual tumorR2 Macroscopic residual tumor

a A tumor nodule greater than 3 mm in diameter in perirectal or pericolic adipose tissue without histologic evidence ofa residual lymph note in the nodule is classified as residual lymph node metastases; however, a tumor nodule up to 3 mmin diameter is classified in the T category as discontinuous extension (ie, T3)b Cases not considered R0 (complete resection) if the following are evident: non-en-bloc resection; radial or bowel mar-gin positive for disease; residual lymph node disease; or NX (incomplete resection)From Nelson H, Petrelli N, Carlin A, et al. Guidelines 2000 for colon and rectal surgery. J Natl Cancer Inst 2001;93:8; withpermission.

preoperative scanning of patients with intraperito-neal colon cancer. Although detection of primarycolon lesions is usually made by colonoscopy orbarium enema, the increased use of CT for a varietyof gastrointestinal symptoms is such that the radiol-ogist may be the first to detect CRC based on CTfindings. CT is most useful for detecting metastaticdisease and regional tumor extension. Complica-tions, such as obstruction, perforation, and fistula,

Table 4: Colon cancer 5-year survival rates

Stage %

Stage I 93Stage IIA 85Stage IIB 72Stage IIIA 83Stage IIIB 64Stage IIIC 44Stage IV 8

I, T1,2,N0M0; IIA, T3 N0M0; IIB, T4N0M0; IIIA, T1,2N1M0;IIIB, T3,4N1M0; IIIC, any T, N2M0; IV, any T, any N, M1.Reprinted with the permission of the American CancerSociety, Inc. All rights reserved.

can be readily visualized. CT is also useful in iden-tifying recurrences, evaluating anatomic relation-ships, documenting normal postoperativeanatomy, and confirming the absence of new le-sions during and after therapy. A recent study ofCT in 130 Veterans Administration patients to de-termine clinical use and cost-effectiveness founddisease previously unknown to the surgeon asfollows: demonstration of local extension (9%);demonstration of metastases (15%); unsuspectedvascular abnormalities (10%); second malignancies(4%); and other pathology (13%). The investiga-tors determined that the preoperative scan directlyaided operative planning in 43 (33%) cases; actu-ally altered management (surgery canceled) in sixcases (5%); and led to qualitatively different carein 16%. The sensitivity and specificity for all metas-tases were 75% and 99%, respectively. For liver me-tastases these were 90% and 99%, respectively [61].Furthermore, a cost savings of $24,000 was realizedover 5 years.

Patients undergo scanning using multidetectorscans after the ingestion of oral contrast and the in-travenous injection of an equivalent dose of 45 g of


iodine in the form of low or iso-osmolar contrastmedium. Routine scans are currently acquired at0.625-mm or 1.25-mm slice thickness, during a sin-gle breathhold, in the portal venous phase of liverenhancement and reconstructed axially for viewingat 5-mm or thinner slices. Oral contrast can begiven the night before to opacify the colon, or canbe given rectally, but this is not routinely indicated,because primary disease has usually already beenconfirmed. Reformatted images may also be usedif determined helpful.

The primary lesion, unless sizable, may not beseen unless the colon has been previously cleansed,an unusual scenario. Even in the uncleansed colon

Fig. 1. Double-contrast barium enema reveals a pe-dunculated left colon polyp with a long stalk.

on routine CT, polypoid or annular lesions can bewell-appreciated because of their enhancementand more solid appearance compared with stool(Fig. 9). Associated findings, such as lymphadenop-athy, peritoneal implants (Fig. 10), tumor penetra-tion through the bowel wall, and colonicobstruction, can be well-appreciated. Tumor ap-pearances may vary from a discrete mass narrowingthe lumen, to bowel wall thickening (see Fig. 10), toa necrotic mass appearing much like an abscess. Incases of associated inflammation and microperfora-tion, the primary differential diagnosis is perforateddiverticulitis. The presence of lymph nodes mayhelp distinguish tumors from diverticulitis, whereasmany other findings are shared by both [62]. Tu-mors may intussuscept and be easily recognizedin longitudinal (Fig. 11) or axial plane. Mucinoustumors may be quite bulky. If mucinous colonicor appendiceal tumors perforate, patients may pres-ent with pseudomyxoma peritonei at CT (Fig. 12).Pericolic tumor extension can be suggested (T3 dis-ease) on CT when the fat planes are blurred, but thisappearance is not specific, nor is it sensitive. Inflam-mation or deep ulceration may cause blurred fatplanes. Normal fat planes may be seen with micro-scopic penetration of the muscularis propria.

Peritoneal surfaces may be involved with tumorin up to 10% to 15% of patients at the time of diag-nosis of CRC, and 40% to 70% of patients with re-current CRC [63]. Although CT may currently bethe best modality to detect early disease, it remainslimited, even with helical technology and thin sli-ces. CRC peritoneal tumor nodules less than 1 cmwere detected in 9% to 24% of 25 patients in a re-cent study in which non–picture archiving andcommunications system observation methods

Fig. 2. (A) Double-contrast barium enema reveals a ring-like density etched-in-white in the transverse colon rep-resenting a polyp viewed en face (arrow). (B) More oblique view reveals a small pedunculated polyp (arrow).

Gollub et al96

were used. Overall, for all sizes, two radiologists de-tected only 60% to 76% of peritoneal implants witha poor interobserver agreement level. Surprisingly,these poor results are little improved from the pre-helical CT era where up to one third of peritonealmetastases were missed in patients undergoing stag-ing laparoscopy for gastric adenocarcinoma [64]. Itis expected that with the use of multidetector scan-ners and picture archiving and communicationssystem for reviewing images, greater accuracy canbe achieved.

CT provides the best resolution of any known im-aging modality and is the first choice for lung nod-ule detection. Despite its high sensitivity, specificityis quite poor. In a recent series of high-risk onco-logic patients with nodules 3 cm or less (75%

Fig. 3. A fecal filling defect (arrow) mimics a polyp inthe cecum (confirmed at colonoscopy).

Fig. 4. Large filling defect in sigmoid colon with a typ-ical lacy, reticulated surface pattern, representinga villous adenoma.

were equal or less than 1 cm), in only 60% ofpatients with a solitary nodule and in only 64%of patients with multiple nodules were the nodulesmalignant [65].

CRC may recur in between 37% and 44% of pa-tients after curative resection, usually within 2 years.Local recurrence accounts for 19% to 48%, whereasdistant metastases account for 25% to 44%. Multi-ple sites of recurrence are most common and localand distant recurrences are more common in rectaltumors. Local recurrences most often occur in theperianastomotic tissues or lymph nodes and maynot be appreciated with luminal examinations likecolonoscopy or barium enema. A recent study of re-current CRC by CTC found that 46 of 51 local recur-rences were in the extraluminal soft tissues [66]. It iscritical to optimize the ability of CT to detect recur-rence at anastomoses and perianastomotic tissuesby ensuring a well-distended colon. CT is in-valuable in assessing the response to chemotherapythrough measurement of index lesions in the lung,liver, lymph nodes, or peritoneum.

Hepatic metastatic diseaseCT scanning performed with the newer multidetec-tor scanners may be timed to obtain images of theliver during both the hepatic arterial phase andthe portal venous phase of hepatic parenchymalenhancement. This may improve both lesion visu-alization and characterization. The use of multide-tector CT scanning with isotropic voxels allows forimages to be reconstructed in any plane in bothphases of contrast administration. For imaging ofhepatic metastases, this generally permits better seg-mental visualization of the metastases relative to

Fig. 5. Double-contrast barium enema reveals an an-nular or apple-core–type appearance indicating pri-mary colon cancer (arrow).


Fig. 6. (A) In the splenic flexure there is an abnormality that is not clearly characterized because of overlappingsegments (arrows). (B) Rotating the patient reveals the pathognomonic appearance for a primary neoplasm.

hepatic arterial and portal venous structures. Thismay aid the surgeon for preoperative planning ofresection of liver metastases [67,68]. The appear-ance of colorectal metastases to the liver dependsin large part on both the phase of contrast adminis-tration in which the scanning is performed and theoverall vascularity of the tumor. In general, colorec-tal metastases appear on CT scan as well-definedareas of low attenuation compared with normal he-patic parenchyma in the portal venous phase ofenhancement (Fig. 13A). In the arterial phase,colorectal metastases may show rim enhancementwith a relatively hyperdense rim. Some larger colo-rectal metastases may demonstrate central areas oflow attenuation likely representing necrosis or cys-tic change. Hepatic metastases may be associatedwith calcification or the development of

calcification [69]. Overall, in a meta-analysis, theCT sensitivity for hepatic metastases was reportedas 72% based on analysis in 1747 patients in 25publications [70]. Overall CT has been shown tohave a detection rate of 85.1% with a positive pre-dictive value of 96.1% and a false-positive rate of3.9%. Some entities that may be confused withcolorectal metastases include hemangioendothelio-ma, hemangioma, hepatic peliosis, biliary ade-noma, biliary hamartoma, and periportal fibrosis[71,72].

CT may also be performed during arterial portog-raphy. Before the advances in multidetector CT andMR imaging this technique was considered themost sensitive for lesion detection. CT arterial por-tography, however, is used less routinely with ad-vances in other imaging modalities and because it

Fig. 7. (A) View of rectum with rectal enema catheter tip in place and balloon inflated. There is no apparentabnormality. (B) After removal of the catheter an irregular lesion is revealed on the lateral wall (arrow). Thisrepresented rectal carcinoma.

Gollub et al98

is an invasive and costly procedure. It remains anoption in the imaging armamentarium, especiallyfor problem-solving cases or in difficult patients.

Virtual colonoscopy

This CT examination (also known as CT colonogra-phy) of the air-filled colon was first described byVining and coworkers [73] in 1994 as a result ofnewly available software, which allowed perspectivevolume rendering (fly-through) of CT data for thefirst time. Since that time, there has been rapid de-velopment and improvement in the software, CTscanner hardware, computer processing speed andstorage capabilities; the understanding of coloniccleansing preparations; reader-strategies and readerexperience; and greater public awareness of the ex-tent and importance of detecting colon polypsand colon cancer. This radiologic test has

Fig. 8. In the ascending colon, near the ileocecal valvelocation, an annular mass is noted (arrows). This rep-resented gradual circumferential invasion from an ex-trinsic ovarian peritoneal implant in the colic gutter.

Fig. 9. Axial contrast-enhanced CT reveals a transversecolon mass (arrow) with wall thickening and narrow-ing of the lumen, consistent with tumor.

contributed greatly the public’s awareness of radiol-ogists and their role in cancer care.

Although created with the ultimate goal of becom-ing a more acceptable screening test for CRC, currentindications, based on extensive evidence-basedresearch, include a superior alternative to bariumenema for total colon examination [74], especiallyin the setting of incomplete colonoscopy [75]; eval-uation of the colon proximal to an obstructingneoplasm [76]; and as an alternative total colonexamination in patients with comorbid factors pre-cluding conventional colonoscopy (eg, on warfarin,pulmonary fibrosis, or allergic to sedatives) or inthose who refuse this or other screening tests. Ithas also been suggested that it is useful in stagingprimary [77] and recurrent [78] colon cancer.

Fig. 10. Coronal-reformatted image in a differentpatient with a mid-transverse colon primary coloncancer manifested by wall thickening and luminalnarrowing (short arrow). Note also the peritonealinfiltration (long arrow), representing extracolonicmetastasis.

Fig. 11. Axial contrast-enhanced CT reveals a masswithin the colon accompanied by mesenteric fatand vessels indicating an intussuscepting carcinoma.


Advantages of CTC over colonoscopy include itsnoninvasiveness, lack of need for sedation, lowercost, and ability to detect significant extracolonicabnormalities and more accurate lesion location.Its main disadvantage is an inability to obtain bi-opsy material, because it is a noninvasive imagingtest.

Fig. 12. Axial CT scan demonstrating scalloping of theliver surface, peritoneal fluid, and omental low-den-sity mass indicative of pseudomyxoma peritonei ina patient with a perforated appendiceal carcinoma.

Using single breathhold helical thin-slice acquisi-tions (typically 1.25–2.5 cm, pitch 0.9–1.375) afterundergoing an accepted colonic cleansing prepara-tion (preferably a dry preparation rather than a wetpreparation [79]) patients are scanned in prone andsupine positions to redistribute retained fluid ormobile stool particles. Room air or carbon dioxideis insufflated by a physician, nurse, the patientthemselves, or an automated pump until good co-lonic distention is achieved as ascertained withbaseline and repeated low-dose scout topograms(Fig. 14). Carbon dioxide is theorized to be reab-sorbed more rapidly and to be more comfortablefor the patient after the examination is completed(based on older barium enema literature). Auto-mated insufflation seems to allow better colonicinsufflation over manual insufflation of CO2, butto date comfort studies comparing similar adminis-tration methods have not determined if air or CO2

is more acceptable to patients [80]. The use ofbowel relaxants (glucagon in the United Statesand hyoscine-N-Butylbromide in Europe) hasbeen suggested to improve distention and patientcomfort, with some investigations showing no im-provement [81] and others suggesting improveddistention and greater patient comfort [82].

Fig. 13. (A) CT demonstrating a left lobe colorectal metastasis. Note on CT its relationship to the hepatic vein. (B)MR image demonstrating a left lobe colorectal metastasis. Note hyperintensity of fat-suppressed T2. (C) MR im-age demonstrating a left lobe colorectal metastasis. Note heterogeneous early enhancement postgadolinium.(D) MR image demonstrating a left lobe colorectal metastasis Note heterogeneous delayed enhancementpostgadolinium.

Gollub et al100

Raw data are reconstructed with 50% to 60%overlap for two-dimensional reformatted imagesin sagittal and coronal planes and for high-qualityvirtual endoscopic views. Using either a primarytwo-dimensional, confirmed with three-dimen-sional, or primary three-dimensional (vice versa)type reading strategy [37], including two-dimen-sional axial images viewed at lung window level,the entire colon is perused in both supine andprone positions for the presence of polyps ormasses. Recently, fecal tagging has been developedto overcome the pitfall of stool particles appearingas polyps, because these do not always contain airor remain mobile [83]. Novel visualization tech-niques, such as virtual dissection views (Fig. 15),are undergoing continued investigation to deter-mine the most accurate and time-efficient modeof interpreting results [84]. Computer-assisted de-tection is also being applied to data sets to improveaccuracy and reader efficiency with some success[85].

Numerous issues are evolving as the technologyand approach undergo changes. Radiation-expo-sure can be quite low in a screening setting becauseof the intrinsic contrast between air and soft tissue.Typical doses for CTC are between 3.6 and 12.2mSv for both acquisitions [86] with microamperageas low as 10 mA being reported to produce diagnos-tic images [87]. Patient comfort studies comparedwith colonoscopy have varied, with generallygood results in favor of high acceptance for CTC[88,89]. Reimbursement by the Center for Medicareand Medicaid Services continues to evolve based onevidence in the literature. Coverage varies with local

Fig. 14. Supine CT scout tomograph after rectal CO2

automated insufflation reveals a gas-density well-outlined colon.

carriers, but has been good in the setting of incom-plete colonoscopy. Screening is not covered, andprobably awaits addition of CTC to the ACS screen-ing recommendations list. Several HMOs have,however, begun to cover for screening CTC basedon aggressive local lobbying and evidence-basedpublications [90].

Finally, an international symposium meets eachyear in Boston to review and present developmentsin CTC and in addition has formed a consortium ofexperts with recommendations for reader training;standards of examination performance (see alsoACR standards [91]); and a new CT ColonographyReporting and Data System coding assignmentwith recommended follow-up intervals [92]. Al-though this system has yet to be validated withlong-term studies, it is a good start to standardizingcommunication between radiologists, clinical col-leagues, and patients.

Filling defects in the colon may represent polyps(Fig. 16); masses (Fig. 17); stool; inverted divertic-ula; mucosal; mural or extramural masses; or nor-mal structures, such as the ileocecal valve orcomplex folds [93]. The two-dimensional images,when adequately thin, demonstrate air in stooland particle mobility between prone and supineimages. Furthermore, a smooth round or lobulatedmorphology can be seen in both stool and polyps,but geometric or irregularly angled borders almostalways represents stool.

Many pitfalls exist, such as desiccated and adher-ent stool particles. The three-dimensional imagesare useful to distinguish true polyps from folds(Fig. 18), but do not show surface details to allowdistinction of stool from polyps (Fig. 19), otherthan morphologic observations, such as acute

Fig. 15. Virtual dissection view of entire colon visual-izing 45 degrees of the wall. Note two larger polyps(arrows) in this patient with familial adenomatouspolyposis.


Fig. 16. (A) Prone axial CT colonography images (viewed at lung-window settings) demonstrating a left colonpolyp (arrow). (B) Supine axial CT colonography images (viewed at lung-window settings) demonstratinga left colon polyp (arrow).

angles, not characteristic of polyps. Flat polyps maybe harder to detect with CTC if the wall of the bowelis not observed at soft tissue windows, because us-ing lung-type windows the reader is relying on grossprotrusions into the air-filled lumen (Fig. 20).

The main limitation to detection of polyps andmasses in the colon is retained stool or fluid, as isthe case for other investigations of the colon, albeitsome of this material can be suctioned at colono-scopy. The development of fecal-tagging followedby electronic subtraction has begun to overcomethis limitation. Here, oral barium and water-solublecontrast agents are added to the colonic-cleansingregimen in varying amounts to label stool and re-tained liquid. Based on higher attenuation at CT,these can be electronically subtracted using special-ized software, and underlying polyps and massescan be discovered (Fig. 21). Another strategy toovercome a poor colon preparation is the additionof intravenous contrast to enhance polyps andmasses and make their presence more conspicuous

Fig. 17. Supine axial CT colonographic image with in-travenous contrast, viewed at soft tissue window set-ting, reveals an annular sigmoid carcinoma (arrow).

in the presence of nonenhancing stool or copiousfluid [94]. Efforts are underway to develop more ac-ceptable, noncathartic colon preparations (or nocleansing preparation at all [95]) with fecal taggingcombined with electronic stool subtraction to en-hance patient compliance with screening for CRC[95,96]. As reader experience improves, colon-cleansing preparations become more tolerable,and a greater number of informed patients seek co-lorectal screening, CTC will likely play an expandingrole in the diagnosis of colon cancer, and more im-portantly in the detection of precancerous polyps.

MR imaging can also be used for virtual colono-scopy with either a bright-lumen technique usingrectally administered water spiked with a paramag-netic contrast agent and imaging with T1-weightedsequences [97], or with a dark-lumen techniqueusing rectal water or air alone and intravenousparamagnetic contrast, using ultrafast three-dimen-sional gradient-echo acquisitions [98]. A recentsmall study showed this technique could diagnoseall lesions 10 mm and greater and 16 of 18 lesions6 to 9 mm in size compared with colonoscopy. Nolesions 5 mm or less were able to be seen [99].Clearly, in younger patients who would receivefrequent surveillance for ulcerative colitis or famil-ial polyposis, the lack of ionizing radiation withthis technique offers an intriguing advantage, butlarger studies with cost-effectiveness analyses areawaited.

UltrasoundTransabdominal ultrasound is the mainstay of eval-uation of the liver in many countries throughoutthe world. Ultrasound evaluation affords a low costtool for evaluation and screening of patients withsuspected liver metastases. At ultrasound, most he-patic metastases are hypoechoic. Some may haveareas of other sonographic patterns including

Gollub et al102

Fig. 18. (A) Prone axial image from CT colonography examination viewed at lung window setting reveals an ap-parent filling defect at the splenic flexure (arrowhead), caused by doubling up of the wall at the inner point ofthe fold to create a pseudomass. (B) Virtual colonoscopic (three dimensional) view pointed at abnormality re-veals that it is a fold right at the point of turn (arrow). (C) Sagittal reformatted image, viewed at lung-windowsetting, clearly depicts location at splenic flexure of colon (arrow).

Fig. 19. Axial supine image from CT colonography,viewed at lung-window settings, shows that anapparent mass at virtual colonoscopy contains air(arrow), revealing its fecal nature.

cystic, calcific, or mixed echogenic patterns. Themost common pattern of metastases to the liver isthat of multiple lesions with hypoechoic halos.Some other metastases to the liver may be echogen-ic, depending on the tumor’s vasculature and ori-gin. Transabdominal ultrasound generally hasa relatively low sensitivity for the detection of livermetastases [100]. It is for this reason that it is cur-rently not an accepted tool for preoperative plan-ning for liver surgery but may be used as anadjunct to CT and MR imaging to evaluate the liverand especially to look at the vasculature and biliarytree. Several new technical developments in ultra-sound including power Doppler and microbubblecontrast agents have improved both detection andcharacterization of solid liver lesions [101]. Con-trast-enhanced ultrasound may be useful for differ-entiating hepatic tumors, especially malignant frombenign ones by their differing enhancementpatterns.


Fig. 20. (A) Axial supine image from CT colonography shows a small filling defect in splenic flexure (arrow), rep-resenting a flat adenoma. (B) Axial prone image of same area reveals lack of movement of same filling defect(arrow). (C) Virtual colonoscopic view of flat adenoma adjacent to a fold (arrow).

MR imagingMR imaging is an imaging technique that usesstrong magnetic fields and radiofrequency pulsesto create an image with outstanding spatial resolu-tion and tissue contrast. Certain nuclei have a mag-netic moment because they are composed of an oddnumber of protons and neutrons. When placed ina strong magnetic field, these nuclei align with thefield and rotate, or spin (precess) about the axis ofthe magnetic field. The frequency of precession(the Larmor frequency) depends on the specific nu-clei and the field strength of the magnet.

Most commonly, the hydrogen nucleus is imagedbecause of its abundance in the body and its highgyromagnetic ratio compared with other nuclei.When hydrogen nuclei in a strong magnetic fieldare excited by the addition of a radiofrequencypulse, these nuclei gain energy. When the radiofre-quency pulse is turned off, the nuclei return to theirresting state and emit the previously absorbed en-ergy at the same frequency. The magnitude of theemitted signal and the time it takes for the nuclei

to return to the resting state depend on certain in-trinsic properties of the nuclei, which include thenuclear spin density (or proton density); longitudi-nal relaxation; transverse relaxation; and flow. Thelongitudinal (or T1) relaxation time is a measureof the time of the precessing nuclei to return to theirbaseline state (ie, oriented parallel to the magneticfield) after the radiofrequency pulse has beenturned off. The transverse (or T2) relaxation time isa measure of the loss of signal in the plane orthog-onal to the long axis of the magnetic field becauseof the loss of phase coherence between the protons.

Recent advancements in MR imaging hardware,software, and contrast agents have made a majorimpact on imaging of the liver. Conventional spinecho pulse sequences have been replaced by fastersequences. The use of the phased-array surface coiltechnology allows for images to be acquired withhigher signal-to-noise, larger matrices, smallerfields of view, and thinner slices. Because of thesetechnologies, almost all liver imaging is now donein a breathhold fashion.

Gollub et al104

Fig. 21. (A) Sagittal reformatted image of right colon from CT colonography performed after fecal tagging.High-attenuation fluid labeled by administration of meglumine diatrizoate obscures underlying colon lumenand wall (arrow). (B) After electronic computer subtraction, air density is restored to this region revealing anunsubtracted, unlabeled (true) abnormality characterized by a small stalk and polyp head (arrow). (C) Virtualcolonoscopic view of small colonoscopically proved pedunculated polyp (arrow). (From Pickhardt PJ, Choi JH.Electronic cleansing and stool tagging in CT colonography: advantages and pitfalls with primary three-dimen-sional evaluation. AJR Am J Roentgenol 2003;181:799–805; with permission.)

Contrast enhancement is also routinely used toevaluate for colorectal metastases to the liver. Ga-dolinium-diethylenetriamine pentaacetic acid isused routinely to improve hepatic lesion detectionand characterization. Other liver-specific contrastagents have been recently introduced further toimprove hepatic lesion detection and characteriza-tion. For instance, mangafodipir trisodium is a he-patocyte-selective contrast agent that is taken upby normal hepatocytes. Normal liver parenchymashows increased signal on T1-weighted post–man-gafodipir trisodium images. Superparamagneticiron oxide is another liver contrast agent, but itis specific for the reticuloendothelial system.Iron oxide causes signal loss within the liver onT2-weighted images. Liver metastases do not takeup the iron oxide; there is an increase in the con-trast of the metastases relative to the liver. The lit-erature is replete with studies comparing allmodalities for evaluation of liver metastases. Forinstance, in one prospective staging study

comparing CT, CT arterial portography, and MRimaging, the sensitivity and specificity for MR im-aging was 82% and 93%, respectively, which wasnot statistically better than CT and CT arterial por-tography [102].

Hepatic metastases may have many different ap-pearances on MR images. On T2-weighted images,most metastases are hyperintense relative to normalhepatic parenchyma (see Fig. 13B). They may be ho-mogeneous or heterogeneous. On T1-weighted im-ages, most hepatic metastases are hypointense.Metastases tend to have irregular borders and maydemonstrate a halo or ring with central hyperinten-sity and lower signal around the periphery. This halopattern has been attributed to peritumoral edema orperitumoral tumor infiltration. Following gadoli-nium enhancement, metastases show heteroge-neous uptake of contrast (Fig. 13C, D). Classically,rim-enhancement with peripheral washout is seenwith dynamic gadolinium-enhanced MR imaging[103,104].


In terms of the value of MR imaging of the liver instaging patients for hepatic resection, a recent studyhas concluded that MR imaging detected not onlyall resectable cases as determined by other imagingmodalities, but three additional cases that were notdeemed resectable with other modalities. The realquestion that remains is, what is the true medicaland economic value of the additional identificationof a subset of patients who can potentially go on tosurgical resection [105]. The relative value versuscost must be carefully considered. Overall, MRimaging of the liver plays a major role in lesioncharacterization and problem solving. It is also im-portant in hepatobiliary imaging for staging andpotentially assessing response to therapy.

Positron emission tomographyPET refers to the acquisition of anatomic and met-abolic data by a PET camera. The most commonlyused radiotracer used for PET imaging is 18-F fluo-rodeoxyglucose (FDG). FDG is an analogue of glu-cose that is transported by glucose transporters andphosphorylated by hexokinase to FDG-6-phos-phate. The polar nature of FDG-6-phosphate re-duces the capacity of the molecule to diffuse outof a cell and the stereochemistry of FDG is suchthat it is not a substrate for any further metabolism.The net effect is a quantifiable, objective measureof the rate of metabolism from the time of injectionto the time of imaging. In general, cancer cells havean accelerated glucose metabolism [106]. Many ofthe pathophysiologic and cellular processes thatlead to or result from inflammation are mimickedby oncogenesis. FDG imaging for cancer is renderedless specific because both inflammation and onco-genesis are associated with accelerated glucose me-tabolism. Newer agents are being investigated butneed Food and Drug Administration approval.

One distinct advantage of PET imaging over ana-tomic imaging is that quantitative results are gener-ated; this allows assessment and reassessment oflocoregional metabolism before and after treatment(the hypothesis being that measurable alterationsin FDG metabolism reflect the biologic processesinduced by a treatment, before a structural changeoccurs). Reliable imaging with FDG requires consis-tency and use of strict practice guidelines to ensuredependability and reproducibility. A large amountof data is acquired and quantitative results are gen-erated. Examples of good practice guidelinesinclude injection of adequate amounts of radioac-tivity and imaging for a sufficient period of timeto ensure images are of sufficient quality for inter-pretation and analysis. If semiquantitative measuresof FDG uptake are to be used for patient follow-up,such as the standardized uptake value, the samemeasure (eg, maximum uptake within a region of

interest versus average uptake) should be used.Consistency in postinjection image acquisitiontimes for subsequent FDG PET scans is necessaryto generate biologically relevant semiquantitativeinformation; the same can be said for the use ofthe same camera with the same reconstruction algo-rithm. Strict adherence to institutional standards forpatient elevated blood glucose (eg, >150 mg/mL) isrecommended so that rescheduling of patients isdone if proper levels do not exist.

FDG PET imaging has a role in the preoperativeevaluation of patients with a biologically aggressiveprimary and in patients with suspected recurrenceto exclude unexpected distant metastases. There isalso a role of FDG PET in the work-up of patientswith occult disease manifesting solely as an elevatedCEA level. The early FDG PET literature was remark-ably consistent across many tumor types wherebythe addition of FDG imaging to conventional imag-ing led to management change in about one thirdof all cases. There have been significant changes inthe technology of cross-sectional imaging devicesin the last 10 years with the widespread implemen-tation of multislice CT scanners and higher fieldstrength MR imaging machines. The increment inPET acquisition technology has been comparativelysmall over the interim with modest changes in sen-sitivity and resolution of PET devices. It is likely thatthe clinical impact of the addition of PET data alonehas lessened as the performance of other modalitieshas improved. The greatest innovation in PET imag-ing devices has not been the PET device itself butrather the successful deployment of PET-CT scan-ners that are fully integrated allowing generationof anatomically coregistered or fused images. Al-though difficult to quantify, it is generally acceptedthat anatomically coregistered imaging improvesthe interpretation of both the PET and CT data. Arepresentative case in which a peritoneal lesionwas missed on CT (Fig. 22A) reveals that it was de-tected on FDG fusion images (Fig. 22B). Finally, animportant and emerging role exists for PET imagingin treatment monitoring.

A further extension of this is to use FDG PET im-aging at an early time point in the course of treat-ment to determine if otherwise toxic therapy isresulting in an expected fall in glucose metabolism.This technique has been used in other tumor typesand parameters associated with good outcomeshave been described [107–109].

A classic indication for CRC imaging with FDGPET is CEA elevation with a negative CT examina-tion. It has been shown that there is a correlationbetween metabolically active tumor bulk as mea-sured by PET and CEA level [110]. As more andmore CT scans are performed using multislice heli-cal technology, the false-negative rate on CT

Gollub et al106

Fig. 22. (A) Axial contrast-enhanced CT scan showing a peritoneal implant (arrow) mimicking unfilled intestine.(B) PET-CT fusion image four-quadrant screen revealing coronal (upper left) and axial (upper right) PET, and un-fused CT (lower left) and fused PET-CT (lower right) with grid lines indicating FDG avid peritoneal implantmissed on CT (see A).

imaging is falling, in part because images are ac-quired faster reducing motion artifacts, and thinnerslices are used, increasing sensitivity for small tu-mor deposits. Isolated CEA elevation with a negativeconventional (CT) imaging study is now a rare sce-nario. A report of 272 cases of CRC that underwentFDG PET imaging found only 15 such cases, ofwhich 14 had true-positive FDG scans, and onewas false-positive. There were four patients whohad symptoms, a negative work-up, and a normalCEA level, and four of whom had cases of positiveFDG PET scans, of which three were false-positive

[111]. Resection of CRC metastases with or withoutadjuvant hepatic arterial pump therapy has led toextraordinary success with long-term follow-updemonstrating up to 60% 10-years survival [112].Correct selection of patients for local therapy is de-pendent on adequate preoperative imaging, ofwhich FDG PET plays an important role. Early stud-ies suggested approximately 30% to 40% of patientswould have their management altered on the basisof a preoperative FDG PET scan [113]. These num-bers seem similar today [114]. The major roleFDG plays in the staging of these patients is in the


identification of extrahepatic disease, rather thanstaging the liver itself. In those patients receivingchemotherapy within 6 months of surgery, the au-thors found that FDG uptake was the strongest pre-dictor of postoperative survival in patients who hadliver resections for CRC [115].

Research continues at a rapid pace into further re-fining measures of metabolic activity and into theuse of more tissue or disease-specific radionuclides.Investigations such as these have opened up a wholenew exciting field of molecular imaging, of whichPET leads the way into the future with the goal ofmore accurate and earlier disease detection andtreatment monitoring.

Newer imaging methods

Exciting advances are occurring in colonic imaging,mainly in the setting of endoscopic visualization.These have been developed with the knowledgethat up to 27% of small adenomas remain undiag-nosed at colonoscopy [30,116]. Chromoendoscopyin combination with high-magnification colono-scopy for in vivo prediction of histology usinga pit-pattern technique has been successful. Near-infrared spectroscopy (Raman effect) during endos-copy can allow distinction of hyperplastic fromadenomatous polyps. Optical coherence tomogra-phy uses light waves to obtain two-dimensionalcross-sectional images of the layers of the gastroin-testinal tract. Based on light scattering this candistinguish adenomas from hyperplastic polyps.Finally, confocal laser microscopy can now be com-bined with videoendoscopy using fiber bundletechnology, and after the injection of fluorescein so-dium has been shown to predict the presence of co-lonic neoplasia with high accuracy [117]. Clearlythese techniques are in their infancy, but they mayprovide more histospecific imaging of CRC, albeiton a submacroscopic or microscopic level beyondthe confines of the specialty of radiology.

Primary and recurrent rectal cancer

Because of its extraperitoneal location, close prox-imity to other organs, higher rate of recurrence,higher morbidity, amenability to different treat-ment strategies, and requirement for subspecializedsurgery, rectal cancer is addressed separately fromcolon cancer in this section.

Approximately 45,000 cases of rectal cancer arediagnosed each year, representing about one thirdof new cases of CRC. Recurrent disease is morecommon in the rectum than in the colon and canresult in pain, immobility, and prolonged hospital-ization. Furthermore, a unique surgical challengepresents because of the confined anatomic space,often demanding subspecialization by colorectal

surgeons, for whom expertise in sphincter-preserv-ing operations is a must.

Important structures must be considered in theradiologic and clinical staging of primary and recur-rent disease including the mesorectal fascia; pelvicsidewall muscles (pyriformis, obturator internus);the levator ani muscles (puborectalis, pubococcy-geus, iliococcygeus); the internal and external analsphincter muscles; nerves (sciatic, autonomic andsacral roots); sacrum; genitourinary organs; and il-iac vessels.

The staging of rectal cancer is identical to coloncancer, but surgical treatment differs. For tumors-in-situ and T1 lesions, patients can undergo transa-nal excision. For T2-T3 tumors that do not involvethe anal sphincters or levator ani, patients can un-dergo anterior resection or an abdominoperinealresection if these structures are involved. Pelvic ex-enteration is often needed if T4 disease is present.In patients with a wide pelvis, laparoscopic surgerycan sometimes be performed. The total mesorectalexcision has become the standard operation for cu-rative intent and removes the rectum with its sur-rounding fascia and lymph nodes in an intactpackage. Staging using the tumor, node, metastasissystem has been shown to be somewhat limitedfor prognosis. T3 lesions, which are close to thecircumferential resection margin (CRM, bulkylesions), have a higher risk of recurrence anda three-fold higher death rate than smaller T3lesions [118].

Treatment of rectal surgery differs between north-ern Europe and other parts of the world, includingthe United States. The use of preoperative chemora-diation is currently considered the standard of carefor T3-T4 or node-positive disease because it down-stages disease, decreases local recurrence, andallows sphincter-preserving surgery. In northernEurope, however, a short course of preoperative ra-diation therapy alone is given, because it is believedthat the total mesorectal excision is adequate to pre-vent recurrence in most patients. If a close margin(1 mm by MR imaging) exists between tumor andthe CRM, these higher-risk patients may get neoad-juvant chemotherapy in addition. As such, determi-nation of the CRM is more important for treatmentdecisions in northern Europe. In the United Statesand elsewhere, it is recognized that this can predictprognosis, but this information does not affecttreatment because all patients get chemotherapyand radiation if they are T3-T4 or node-positive.

Most rectal tumors are diagnosed by the digitalrectal examination. It has an accuracy of 67% to83% [119] and gives a surgeon an idea of the loca-tion, bulkiness, and fixation of the tumor, but canlead to understaging in 47% of cases. It providesonly indirect evidence about sphincter infiltration.

Gollub et al108

Fig. 23. Endorectal ultrasound image obtained with a radial probe revealing hypoechoic masses seemingly dis-rupting all layers in two different patients, and representing pathologically proved T2 (A) and T3 disease (B).(Courtesy of W. Douglas Wong, MD, New York, NY.)

Recently, MR imaging has been found to be moreaccurate at predicting an involved CRM comparedwith digital rectal examination with a specificity of84% versus 29% [120].

Endorectal ultrasound (ERUS) is the most estab-lished and preferred modality for T-staging of rectalcancer. It is usually part of the initial digital rectalexamination performed by surgeons, although insome centers it is performed by radiologists. Its ac-curacy is 80% to 90%. It is limited in distinguishingbetween T2 and early T3 disease (Fig. 23), and re-cently, based on data from the authors’ institution,a modification of the staging system has been pro-posed whereby uT2–early uT3 lesions are classifiedas uTy wherein pathologic features and nodal statusdetermine need for neoadjuvant therapy [121]. An-other limitation of ERUS is its limited ability to vi-sualize the mesorectal fascia. Tumor overstagingoccurs in 11% to 18% of cases and understagingin 5% to 13% of cases [121]. For nodal staging,the ERUS accuracy is 64% to 83%. The main limita-tion is the use of size criteria (as for other structuralimaging modalities). One study found that twothirds of metastatic nodes in CRC were <5 mm.Overstaging occurs in 5% to 22% (inflammatoryenlargement) and understaging in 2% to 25% (no-des beyond the range of the probe). At the authors’institution, nodes that are 3 mm or more, round,hypoechoic, and in the appropriate location are re-garded as potentially positive (Fig. 24).

In comparison, even when optimized with rectalcontrast, glucagon, and prone thin-slice imaging,CT is limited for local staging because of its inher-ent low soft tissue contrast (Fig. 25), which doesnot allow for accurate approximation of T stage un-less there is gross invasion of adjacent organs (T4),and even here many false-positive cases are seen(Fig. 26). Errors are usually caused by incorrect di-agnosis of pyriform muscle, pelvic floor, or sacralinvasion. A meta-analysis of 83 published studies

of rectal cancer between 1980 and 1998 comparedthe performance characteristics of ERUS, MR imag-ing, endorectal MR imaging, and CT for T and Nstaging and found that CT performed poorly com-pared with other modalities with a T-stage sensitiv-ity of 78%, specificity of 63%, and accuracy of 73%compared with ERUS, the most accurate (sensitivityof 93%, specificity of 78%, and accuracy of 87%).CT performed poorly for nodes with a sensitivityof 52%, specificity of 78%, and accuracy of 66%compared with ERUS (sensitivity of 71%, specificityof 76%, and accuracy of 74%). This is likely causedby the reliance on size criteria at CT, whereas at ul-trasound one can use some additional morphologiccriteria, such as shape and echogenicity. T-stage sen-sitivity on CT is even more limited in those studiesin which preoperative radiation was used (probablybecause of blurring of fat planes) [122]. Anothermeta-analysis of 90 studies between 1985 and2002 found no statistically significant differencein the receiver operating characteristic curves be-tween CT, ERUS, and MR imaging for nodal staging,but found superiority of ERUS for perirectal soft

Fig. 24. Endorectal ultrasound image revealing sev-eral rounded hypoechoic lymph nodes in the meso-rectal fascia (arrows). (Courtesy of W. DouglasWong, MD, New York, NY.)


Fig. 25. (A) Supine pelvic CT scan after administration of rectal barium revealing semiannular rectal carcinomaalong right wall. (B) Prone pelvic CT scan in a different patient after administration of rectal air and intravenousglucagon revealing sessile carcinoma at 12- to 2-o’clock position with adjacent mesorectal lymph node enlarge-ment (arrow).

tissue invasion. CT sensitivity for N staging on aver-age was 55% [123]. Little data exist on performancecharacteristics for multidetector helical CT. A recentprospective two-reader study of 55 patients usingfour-row multislice CT found higher average accu-racy’s for T, N, and overall stage compared withnonhelical studies. Coronal and sagittal two-di-mensional reformatted images improved perfor-mance even more. For example, T-stage accuracyincreased from 81%/77% to 98%/90% (P 5 .02)for two readers; N-stage accuracy increased from73%/71% to 96%/80% (P 5 .01); and overall clin-ical stage accuracy also increased significantly[124]. N staging is particularly difficult after chemo-radiotherapy because of blurring of fat planes. Inthe authors’ study of 78 patients with ERUS T3-T4or N-positive disease who received 5040 Gy and5-fluorouracil–based therapy and underwent pre-therapy and 6-week posttherapy optimizedmultidetector helical CT scans, followed by totalmesorectal excision resection and whole-mount

Fig. 26. Axial CT at the level of the pelvic floor revealsan ovoid ring-enhanced mass lesion arising from therectum in a patient with known rectal primary carci-noma and seemingly inseparable from the prostategland (arrow). After abdominoperineal resectionand partial prostatectomy, no tumor was seen invad-ing the prostate at histopathology.

pathologic specimens, two radiologists in consen-sus achieved an overall accuracy of only 65% in de-termining involved mesorectal lymph nodes, witha good interobserver agreement (k 5 0.64, unpub-lished data).

MR imaging displays superior soft tissue contrastto CT. It has excellent resolution and can depict theentire mesorectum. This is possible using phased-array coils, small fields of view, fast T2-weighted se-quences, stronger magnetic gradients, and thinnersections. Nonetheless, limitations do exist for dis-tinction between T2 and early T3 lesions, becausestranding into the mesorectal fat may represent in-flammation and be overstaged as T3 or microscopictumor extending into this fat may not be seen andbe understaged as T2. Furthermore, nodal stagingis limited because of the use of size criteria as forother morphologic imaging modalities. In theKwok meta-analysis, overall T-stage accuracy was

Fig. 27. Sagittal fast spin echo (FSE) T2-weighted pel-vic MR image using pelvic 8-channel phased-array coil(TR/TE of 470/107, NEX 5 1, 4-mm thickness, matrix256 � 192) revealing a large mass of the rectum in-vading the sacrum (arrows).

Gollub et al110

82% with a sensitivity of 86% and a specificity of77%. For N staging the sensitivity was 76%, speci-ficity 80%, and accuracy 74%. In publications thatused endorectal coil MR imaging, accuracy’s werehigher, but limitations still exist in coil designsuch that stenotic or proximal tumors cannot al-ways be easily imaged. Cases of overstaging andunderstaging were seen in from 6% to 13% forboth phased-array and endorectal coil MR imaging.

Recent data support MR imaging as an overall su-perior modality to stage rectal cancer. Use of newlydefined criteria for nodes by one investigator wasfound to allow N-stage sensitivity of 85% and spec-ificity of 97%. Compared with ERUS and digital rec-tal examination, MR imaging was superior in termsof clinical benefit, cost-effectiveness, assessment ofdepth of invasion, lymph node involvement, andCRM status [125].

Because of the importance of the CRM in deter-mination of treatment in certain European centers,the Magnetic Resonance Imaging and Rectal Cancer

Fig. 28. Axial FSE T2-weighted MR image (TR/TE of500/104, NEX 1, 4-mm thickness, 256 � 256) revealinga bulky low rectal tumor invading anteriorly obliter-ating (arrows) the posterior vaginal muscular wall(low T2 signal).

European Equivalence Study from 11 European cen-ters investigated the use of MR imaging as a preoper-ative tool compared with histopathology for itsability to predict the CRM. Abstracted early data,not as yet comprehensively published, indicated95% accuracy in predicting a clear CRM (tumorgreater than 1 mm from the CRM) compared withhistopathology. MR imaging was equivalently accu-rate in determining tumor depth compared withhistopathology to within an error of 0.5 mm [126].

At the authors’ institution, an eight-channelphased array coil is used with a 1.5-T magnet. Theauthors perform fast spin echo T2-weighted axial,oblique axial (angled perpendicular to the rectal lu-men and sacrum), sagittal, and coronal images andT1-weighted axial fat-saturation images and sagittalT1-weighted gradient echo type gadolinium-enhanced perfusion images. Local staging of pri-mary tumor can be well depicted with small field ofview, 512 � 512 matrix images, using several signalaverages, not requiring breathholding (Figs. 27–31).

Recurrent rectal cancer occurs in 4% to 30% ofpatients and is isolated in 25% to 50%. It can be dif-ficult to detect with CT and even on clinical exami-nation. It often involves major structures. Survivalis usually 3 to 6 months without treatment. Surgeryis advised to avert morbidity and prolong survivalbut only if an R0 resection (no gross or microscopicresidual disease) can be achieved [127]. MR imag-ing is superior to CT to predict invasion (sensitivity97% versus 70%), but both can falsely predict mus-cle invasion. The distinction between tumor andscar tissue related to recent surgery or radiationmay be quite difficult and can often be aided byPET imaging (see later). Although MR imagingcan distinguish between muscle and tumor usinggadolinium enhancement differences, errors canbe caused by briskly enhancing scar (immature fi-brosis) within 1 year of surgery. Furthermore, low

Fig. 29. (A) Oblique axial FSE T2-weighted pelvic MR image through mid-pelvis demonstrating rectal tumor ex-tending into perirectal soft tissues and invading cervical stroma (dark T2 ring, arrow). (B) Standard axial FSE T2-weighted MR image at same level reveals uncertain relationship with cervix, seen now in obliquity.


T2-weighted signal masses do not always representpure fibrosis. At histologic examination, biopsiesof low T2 areas have also contained malignant cells.Tumor may also be desmoplastic, and atypicallylow in signal on T2-weighted sequences [128]. Ina study of 139 rectal MR imaging examinations atthe authors’ institution, 46 patients had localizeddisease and 33 had pathologic correlation. A totalof 133 of 156 areas of recurrence were agreed onby two radiologists (85% concordance). In 77 path-ologically proved involved organs at curative resec-tion in 33 patients, using MR imaging the authorsidentified 72 (93.5%) and missed 5 (6.5%). The

Fig. 30. Axial FSE T2-weighted pelvic MR image show-ing bulky rectal tumor invading and obliterating nor-mal high T2 signal fluid appearance of seminalvesicles (arrows).

Fig. 31. Axial FSE T2-weighted pelvic MR image re-veals a large intermediate signal rectal mass with ex-tension to the peripheral zone of the prostate(arrows).

false-positive rate was 15.58% and the false-nega-tive rate was 4.58% (Figs. 32–34) (LawrenceSchwartz, unpublished data). Prediction of inva-sion of the bladder, prostate, and sacral periosteumwhen not grossly obvious can be limited. Seminalvesicle invasion may be the easiest to predict be-cause of replacement of high-signal T2 fluid by tu-mor (Fig. 35).

Positron emission tomographyin rectal cancer

Low rectal cancers represent a particular patientpopulation that may behave differently from the re-mainder of patients with CRC. The low rectum haslymphatic drainage to inguinal nodes and the morecommon pelvic side wall and intra-abdominal no-des. The venous drainage of the low rectum may by-pass the liver, and lead to increase in pulmonarymetastases in this group. For these very reasons a re-cent investigation found that FDG PET-CT alteredtreatment plans in patients with untreated rectalcancers 38% of the time primarily by detecting dis-ease in inguinal nodes [129] with most occult dis-ease found in tumors within 6 cm of the analverge. A case of rectal cancer with inguinal nodesis presented in Fig. 36.

Induction chemotherapy or induction chemora-diation therapy is often given to patients with ad-vanced rectal cancer to downstage disease, allowsphincter-preservation, and decrease local recur-rence rates. A number of groups have investigatedthe use of posttreatment, preoperative FDG PETin the prediction of local control and overall sur-vival. It is hypothesized that a positive FDG scanposttreatment may portend a biologically more

Fig. 32. Coronal FSE T2-weighted pelvic MR image(TR/TE of 4466/105, NEX 1, 4-mm thickness, 256 �192) reveals intermediate signal tumor metastasis in-vading sacrum and sacral nerve roots.

Gollub et al112

Fig. 33. (A) Axial FSE T2-weighted pelvic MR image revealing a mass invading the left sciatic nerve at the greatersciatic foramen (arrow). (B) Coronal FSE T2-weighted pelvic MR image revealing a mass invading the left sciaticnerve (arrow) at the greater sciatic foramen in a different patient.

aggressive or less chemosensitive–type tumor. Incontradistinction, if complete pathologic responsesto therapy can be predicted by PET, surgery mightin theory be avoided. It seems that FDG PET isthe current gold standard in determining the re-sponse to induction treatment [130] and prognosis[131].

Radiologic follow-up and treatmentmonitoring

Surgery is the mainstay of therapy for CRC. Seventypercent to 80% of patients have tumors that can beresected with curative intent. Adjuvant radiationtherapy, chemotherapy, or both are useful in

Fig. 34. (A) Axial FSE T2-weighted pelvic MR image revealing a low-signal infiltrative mass invading the rightureter (arrow). (B) Axial FSE T2-weighted pelvic MR image revealing a low-signal infiltrative mass invadingthe right ureter and right bladder wall (arrowheads). (C) Axial FSE T2-weighted pelvic MR image revealinga low-signal infiltrative mass invading the right pelvic sidewall (arrow). (D) Axial FSE T2-weighted pelvic MR im-age revealing a low-signal infiltrative mass invading the right pelvic sidewall (arrow).


selected patients. Among patients who have under-gone resection for localized disease, the 5-year sur-vival rate is 90%. The rate is only 65% if lymphnode metastases are present. Thirty percent to40% of patients recur within the first 2 to 3 years

Fig. 35. Axial FSE T2-weighted pelvic MR image re-vealing recurrent low-signal tumor in the presacralspace growing anteriorly to invade the left seminalvesicle (arrow).

[132,133]. The most common sites of relapse arethe liver, the local site, the abdomen, and the lung.

Current recommendations for follow-up after re-section include (1) physician visits every 3 to 6months for 3 years with decreasing frequency there-after; (2) colonoscopy every 3 to 5 years dependingon findings, with continued surveillance after 5years; and (3) serum CEA levels every 3 to 6 monthsfor 5 years [133].

CEA elevation is associated with disease recur-rence in 60% to 70% of cases. Most often these recur-rences are found to be unresectable anyway. As such,it has been estimated that if all patients were sub-jected to CEA surveillance, only 0.7% would becured by CEA-directed second-look surgery [132].

Because in part of the lack of evidence-based dataregarding the efficacy of imaging for surveillance,the American Society of Clinical Oncology andthe National Comprehensive Cancer Network donot recommend CT imaging for surveillance,whereas the European Society of Medical Oncologyrecommends liver ultrasound annually for 3 years.

In the first prospective randomized multicentercontrolled clinical trial performed of its type, ithas been shown that more intensive follow-up of

Fig. 36. Fusion PET-CT in low rectal cancer demonstrating sagittal PET showing multiple foci of FDG avidity in therectum (upper left), rectal and right inguinal activity (axial, upper right), nonfused CT (lower left), and fusedPET-CT (lower right) with enlarged inguinal nodes and mildly prominent lower rectum corresponding to areaof tumor.

Gollub et al114

patients with stage II CRC and of rectal tumorsspecifically can lead to higher overall survival com-pared with simple surveillance. ‘‘Intensive follow-up’’ was defined as the addition of CT, ultrasound,and colonoscopy to routine clinical examination,CEA, and other blood tests. More intensive surveil-lance also led to a higher proportion of resectabletumor recurrence [134], but the difference in me-dian survival (50 months versus 35 months) didnot reach statistical significance, possibly becauseof a small sample size. Other studies may havebeen similarly underpowered to detect a true differ-ence between intensive and standard monitoring.Large randomized studies are required to addressthis question.

At several institutions, including our own , oncol-ogists typically obtain annual CT scans of the abdo-men and pelvis for the first 3 years, but this is notbased on any formal assessment of utility. Furtherlarger studies are needed to confirm that this isa cost-effective and clinically effective strategy interms of overall survival and quality of life.

For patients at higher risk, such as those with fa-milial adenomatous polyposis, hereditary nonpoly-posis CRC, or ulcerative colitis, guidelines for morefrequent endoscopy exist, but there are no formalimaging recommendations.

Summary

The role of imaging in the screening, diagnosis,staging, and restaging of CRC is vast, ranging fromdetection of precancerous minute adenomas usingcolonoscopy or virtual colonoscopy to accurate con-firmation and restaging of recurrent disease using theexquisite resolution of MR imaging and the highlysensitive metabolic imaging technique of PET scan-ning. Equally encouraging is the continual evolutionand recreation of new roles for radiology, such asrefinements in lymph node MR imaging with ironoxide agents and the potential to fuse imaging mo-dalities (eg, PET and CTC [135]), which introducesand expands the role that molecular imaging mayplay in diagnosis and perhaps ultimately in therapy.

Although the understanding of the developmentand progression of CRC has come a long way, hugehurdles lay ahead in the ability to access and screenlarge numbers of average-risk individuals. Untilsuch time as that is achieved, this all too prevent-able neoplasm will continue to be a leading cancerkiller.

References

[1] Spann S, Levin B, Rozen P, et al. Colorectal can-cer: How big is the problem, why prevent it andhow might it present? In: Rozen P, Young G,

Levin B, editors. Colorectal cancer in clinicalpractice: prevention, early detection andmanagement. London: Martin Dunitz; 2002.p. 1–13.

[2] Mulhall BP, Veerappan GR, Jackson JL. Meta-analysis: computed tomographic colonography.Ann Intern Med 2005;142:635–50.

[3] American Cancer Society. Cancer facts and figures2006. Atlanta: American Cancer Society; 2006.

[4] Nusko G, Mansmann U, Partzsch U, et al. Inva-sive carcinoma in colorectal adenomas: multi-variate analysis of patient and adenomacharacteristics. Endoscopy 1997;29:626–31.

[5] Gore RM, Gelfand DW, Smith C, et al. Colorec-tal cancer: clinical and pathologic features,screening strategies, radiologic diagnosis and ra-diologic staging. Radiol Clin north Am 1997;35:403–87.

[6] Goldberg RM, Sargent DJ, Morton RF, et al. Arandomized controlled trial of fluorouracilplus leucovorin, irinotecan and oxaliplatincombinations in patients with previously un-treated metastatic colorectal cancer. J Clin On-col 2004;1:23–30.

[7] Pignone M, Saha S, Hoerger T, et al. Cost-effec-tiveness analysis of colorectal cancer screening:a systematic review for the US Preventive ServicesTask Force. Ann Intern Med 2002;137:96–104.

[8] Imperiale TF, Ransohoff DF, Itzkowitz SH, et al.Fecal DNA versus fecal occult blood for colorec-tal-cancer screening in an average-risk popula-tion. N Engl J Med 2004;351:2704–14.

[9] Ristvedt SL, McFarland EG, Weinstock LB, et al.Patient preferences for CT colonography, con-ventional colonoscopy, and bowel preparation.Am J Gastroenterol 2003;98:578–85.

[10] Muto T, Bussey HJ, Morson BC. The evolutionof cancer of the colon and rectum. Cancer1975;36:2251–70.

[11] Winawer SJ, Fletcher RF, Rex D, et al. Colorectalcancer screening and surveillance: clinicalguidelines and rationale. Update based on newevidence. Gastroenterology 2003;124:544–60.

[12] Ferrucci JT. Virtual colonoscopy for colon can-cer screening: further reflections on polyps andpolitics. AJR Am J Roentgenol 2003;181:795–7.

[13] Hawk ET, Levin B. Colorectal cancer preven-tion. J Clin Oncol 2005;23:378–91.

[14] Mak T, Lalloo F, Evans DGR, et al. Molecularstool screening for colorectal cancer. Br J Surg2004;91:790–800.

[15] Podolsky DK. Going the distance: the case fortrue colorectal cancer screening. N Engl J Med2000;343:207–8.

[16] Fearon ER, Vogelstein B. A genetic model for co-lorectal tumorigenesis. Cell 1990;61:759–67.

[17] Levine MS, Glick SN, Rubesin SE, et al. Double-contrast barium enema examination and colo-rectal cancer: a plea for radiologic screening.Radiology 2002;222:313–5.

[18] Winawer SJ, Stewart ET, Zauber AG, et al. Acomparison of colonoscopy and double-


contrast barium enema for surveillance afterpolypectomy. N Engl J Med 2000;342:1766–72.

[19] Scheitel SM, Ahlquist DA, Wollan PC, et al. Co-lorectal cancer screening: a community case-control study of proctosigmoidoscopy, bariumenema radiography, and fecal occult bloodtest efficacy. Mayo Clin Proc 1999;74:1207–13.

[20] Rex DK, Rahmani EY, Haseman JH, et al. Rela-tive sensitivity of colonoscopy and bariumenema for detection of colorectal cancer in clin-ical practice. Gastroenterology 1997;112:17–23.

[21] Fletcher RH. The end of barium enemas?[editorial]. N Engl J Med 2000;342:1823–4.

[22] Glick SN. Comparison of colonoscopy anddouble-contrast barium enema, correspon-dence. N Engl J Med 2000;343:1728–30.

[23] Winawer SJ, Fletcher RH, Miller L, et al. Colo-rectal cancer screening: clinical guidelinesand rationale. Gastroenterology 1997;112:594–642.

[24] Citarda F, Tomaselli G, Capocaccia R, et al. Effi-cacy in standard clinical practice of colono-scopic polypectomy in reducing colorectalcancer incidence. Gut 2001;48:812–5.

[25] Winawer SJ, Zauber AG, Ho MN, et al. Preven-tion of colorectal cancer by colonoscopicpolypectomy: The National Polyp Study Work-group. N Engl J Med 1993;329:1977–81.

[26] National Cancer Institute Web site. Available at:http://www.nci.nih.gov/cancertopics/factsheet/ColorectalPLCO.

[27] Lieberman DA, Weiss DG, Bond JH, et al. Use ofcolonoscopy to screen asymptomatic adults forcolorectal cancer. N Engl J Med 2000;343:162–8.

[28] Habr-Gama A, Waye JD. Complications andhazards of gastrointestinal endoscopy. World JSurg 1989;13:193–201.

[29] Blakeborough A, Sheridan MB, Chapman AH.Complications of barium enema examinations:a survey of UK consultant radiologists 1992 to1994. Clin Radiol 1997;52:142–8.

[30] Rex DK, Cutler CS, Lemmel GT, et al. Colono-scopic miss rates of adenomas determined byback-to-back colonoscopies. Gastroenterology1997;112:24–8.

[31] Lieberman D. Quality and colonoscopy: a newimperative. Gastrointest Endosc 2005;61:392–4.

[32] Pickhardt PJ, Nugent PA, Mysliwiec PA, et al.The adenoma miss rate at optical colonoscopy:a novel assessment using virtual colonoscopy asa separate reference standard. Ann Intern Med2004;141:352–9.

[33] Smith RA, Cokkinides V, von Eschenbach AC,et al. American Cancer Society Guidelines forearly detection of cancer. CA Cancer J Clin2002;52:8–22.

[34] Mysliwiec PA, Brown ML, Klabunde CN, et al. Arephysicians doing too much colonoscopy? A na-tional survey of colorectal surveillance after pol-ypectomy. Ann Intern Med 2004;141:264–71.

[35] Seeff LC, Manninen DL, Dong FB, et al. Is thereendoscopic capacity to provide colorectal

cancer screening to the unscreened populationin the United States? Gastroenterology 2004;127:1841–4.

[36] Saitoh Y, Obara T, Watari J, et al. Invasion depthdiagnosis of depressed type early colorectal can-cers by combined use of videoendoscopy andchromoendoscopy. Gastrointest Endosc 1998;48:362–70.

[37] Pickhardt PJ, Choi JRC, Hwang I, et al. Com-puted tomographic virtual colonoscopy toscreen for colorectal neoplasia in asymptomaticadults. N Engl J Med 2003;349:2191–200.

[38] Halligan S, Altman DG, Taylor SA, et al. CT co-lonography in the detection of colorectalpolyps and cancer: systematic review, meta-analysis, and proposed minimum data set forstudy level reporting. Radiology 2005;237:893–904.

[39] Banerjee S, Van Dam J. CT colonography for co-lon cancer screening. Gastrointest Endosc 2006;63:121–33.

[40] Cotton PB, Durkalski VL, Pineau BC, et al.Computed tomographic colonography (virtualcolonoscopy): a multicenter comparison withstandard colonoscopy for detection of colorec-tal neoplasia. JAMA 2004;291:1713–9.

[41] Ferrucci JT. Colonoscopy: virtual and optical-another look, another view. Radiology 2005;235:13–6.

[42] Sonnenberg A, Delco F, Bauerfeind P. Is virtualcolonoscopy a cost-effective option to screenfor colorectal cancer? Am J Gastroenterol1999;94:2268–74.

[43] Ladabaum U, Song K, Fendrick AM. Colorectalneoplasia screening with virtual colonoscopy:when, at what cost, and with what nationalimpact? Clin Gastroenterol Hepatol 2004;2:554–63.

[44] Heitman S, Fong A, Dean S, et al. Cost-effective-ness study of CT colonography versus colonos-copy using decision analysis. [abstract].Gastroenterology 2004;126:A202.

[45] van Gelder RE, Florie J, Stoker J. Colorectal can-cer screening and surveillance with CT colonog-raphy: current controversies and obstacles.Abdom Imaging 2005;30:5–12.

[46] van Gelder RE, Birnie E, Florie J, et al. Patientexperience and preference with CT colonogra-phy and conventional colonoscopy: a five-week follow-up study. Radiology 2004;233:328–37.

[47] Burling D, Halligan S, Slater A, et al. Poten-tially serious adverse events at CT colonogra-phy in symptomatic patients: national surveyof the United Kingdom. Radiology 2006;239:464–71.

[48] Sosna J, Blachar A, Amitai M, et al. Colonic per-foration at CT colonography: assessment of riskin a multicenter large cohort. Radiology 2006;239:457–63.

[49] van Gelder RE, Nio CY, Florie J, et al. Computedtomographic colonography compared with

http://www.nci.nih.gov/cancertopics/factsheet/ColorectalPLCO

http://www.nci.nih.gov/cancertopics/factsheet/ColorectalPLCO

Gollub et al116

colonoscopy in patients at increased risk for co-lorectal cancer. Gastroenterology 2004;127:41–8.

[50] Rubesin SE, Levine MS, Laufer I, et al. Double-contrast barium enema examination technique.Radiology 2000;215:642–50.

[51] Eide TJ. Colorectal polyps: a symposium. WorldJ Surg 1991;15:1–56.

[52] Torlakovic E, Snover DC. Serrated adenomatouspolyposis in humans. Gastroenterology 1996;100:748–55.

[53] Levine MS, Rubesin SE, Laufer I, et al. Diagnosis ofcolorectal neoplasms at double-contrast bariumenema examination. Radiology 2000;216:11–8.

[54] Rembacken BJ, Fujii T, Cairns A, et al. Flat anddepressed colonic neoplasms: a prospectivestudy of 1000 colonoscopies in the UK. Lancet2000;355:1211–4.

[55] Saitoh Y, Waxman I, West AB, et al. Prevalenceand distinctive biologic features of flat colorec-tal adenomas in a North American population.Gastroenterology 2001;120:1657–65.

[56] Fujiya M, Maruyama M. Small depressed neo-plasms of the large bowel: radiographic visuali-zation and clinical significance. AbdomImaging 1997;22:325–31.

[57] Lee JT, Dixon MR, Murrell Z, et al. Colonic his-toplasmosis presenting as colon cancer in thenon immunocompromised patient: report ofa case and review of the literature. Am Surg2004;70:959–63.

[58] Ott DJ, Chen YM, Gelfand DW, et al. Single-contrast vs double-contrast barium enema inthe detection of colonic polyps. AJR Am JRoentgenol 1986;146:993–6.

[59] Prokop M. Image processing and display tech-niques. In: Prokop M, Galnski M, editors. Spiraland multislice computed tomography of thebody. Sttugart: Thieme; 2003. p. 45–82.

[60] Brink JA. Contrast optimization and scan tim-ing for single and multidetector-row computedtomography. J Comput Assist Tomogr 2003;27(Suppl 1):S3–8.

[61] Mauchley DC, Lynge DC, Langdale LA, et al.Clinical utility and cost-effectiveness of routinepreoperative computed tomography scanningin patients with colon cancer. Am J Surg 2005;189:512–7.

[62] Chintapalli KN, Esola CC, Chopra S, et al. Peri-colic mesenteric lymph nodes: an aid in distin-guishing diverticulitis from cancer of the colon.AJR Am J Roentgenol 1997;169:1253–5.

[63] De Bree E, Koops W, Kroger R, et al. Peritonealcarcinomatosis from colorectal or appendicealorigin: correlation of preoperative CT with in-traoperative findings and evaluation of interob-server agreement. J Surg Oncol 2004;86:64–73.

[64] Burke EC, Karpeh MS, Conlon KC, et al. Lapa-roscopy in the management of gastric adenocar-cinoma. Ann Surg 1997;225:262–7.

[65] Ginsberg MS, Griff SK, Go BD, et al. Pulmonarynodules resected at video-assisted thoracoscopic

surgery: etiology in 426 patients. Radiology1999;213:277–82.

[66] You YT, Chien CRC, Wang JY, et al. Evaluation ofcontrast-enhanced computed tomographic colo-nography in detection of local recurrent colorec-tal cancer. World J Gastroenterol 2006;12:123–6.

[67] Fetita C, Lucidarme O, Preteux F, et al. CT he-patic venography: 3D vascular segmentationfor preoperative evaluation. Med Image Com-put Assist Interv Int Conf 2005;8(Pt 2):830–7.

[68] Sahani D, Mehta A, Blake M, et al. Preoperativehepatic vascular evaluation with CT and MR an-giography: implications for surgery. Radio-graphics 2004;24:1367–80.

[69] Easson AM, Barron PT, Cripps C, et al. Calcifica-tion in colorectal hepatic metastases correlateswith longer survival. J Surg Oncol 1996;63:221–5.

[70] Kinkel K, Lu Y, Both M, et al. Detection ofhepatic metastases from cancers of the gastroin-testinal tract by using noninvasive imagingmethods (US, CT, MR imaging, PET): a meta-analysis. Radiology 2002;224:748–56.

[71] Abdelmoumene A, Chevallier P, Chalaron M,et al. Detection of liver metastases under 2cm: comparison of different acquisition proto-cols in four row multidetector-CT (MDCT).Eur Radiol 2005;15:1881–7.

[72] Valls C, Andia E, Sanchez A, et al. Hepatic me-tastases from colorectal cancer: preoperative de-tection and assessment of resectability withhelical CT. Radiology 2001;218:55–60.

[73] Vining DJ, Gelfand DW, Bechtold RE, et al.Technical feasibility of colon imaging with heli-cal CT and virtual reality [abstract]. AJR Am JRoentgenol 1995;162:194.

[74] Johnson CD, MacCarty RL, Welch TJ, et al.Comparison of the relative sensitivity of CT co-lonography and double-contrast barium enemafor screen detection of colorectal polyps. ClinGastroenterol Hepatol 2004;2:314–21.

[75] Macari M, Berman P, Dicker M, et al. Usefulnessof CT colonography in patients with incom-plete colonoscopy. AJR Am J Roentgenol 1999;173:561–4.

[76] Fenlon HM, McAneny DB, Nunes DP, et al. Oc-clusive colon carcinoma: virtual colonoscopy inthe preoperative evaluation of the proximal co-lon. Radiology 1999;210:423–8.

[77] Neri E, Giusti P, Battolla L, et al. Colorectal can-cer: role of CT colonography in preoperativeevaluation after incomplete colonoscopy. Radi-ology 2002;223:615–9.

[78] Fletcher JG, Johnson CD, Krueger WR, et al.Contrast-enhanced CT colonography in recur-rent colorectal carcinoma: feasibility of simulta-neous evaluation for metastatic disease, localrecurrence, and metachronous neoplasia in co-lorectal carcinoma. AJR Am J Roentgenol 2002;178:283–90.

[79] Macari M, Lavelle M, Pedrosa I, et al. Effect of dif-ferent bowel preparations on residual fluid at CTcolonography. Radiology 2001;218:274–7.


[80] Burling D, Taylor SA, Halligan S, et al. Auto-mated insufflation of carbon dioxide forMDCT colonography: distension and patientexperience compared with manual insufflation.AJR Am J Roentgenol 2006;186:96–103.

[81] Yee J, Hung RK, Akerkar GA, et al. The useful-ness of glucagon hydrochloride for colonicdistention in CT colonography. AJR Am J Roent-genol 1999;173:169–72.

[82] Taylor SA, Halligan S, Goh V, et al. Optimizingcolonic distention for multi-detector row CT co-lonography: effect of hyoscine butylbromideand rectal balloon catheter. Radiology 2003;229:99–108.

[83] Lefere PA, Gryspeerdt SS, Dewyspelaere J, et al.Dietary fecal tagging as a cleansing methodbefore CT colonography: initial results polypdetection and patient acceptance. Radiology2002;224:393–403.

[84] Hoppe H, Quattropani C, Spreng A, et al. Vir-tual colon dissection with CT colonographycompared with axial interpretation and conven-tional colonoscopy: preliminary results. AJRAm J Roentgenol 2004;182:1151–8.

[85] Summers RM, Jerebko AK, Franaszek M, et al.Colonic polyps: complementary role of com-puter-aided detection in CT colonography. Ra-diology 2002;225:391–9.

[86] van Gelder RE, Venema HW, Serlie IW, et al. CTcolonography at different radiation dose levels:feasibility of dose reduction. Radiology 2002;224:25–33.

[87] Cohnen M, Vogt C, Beck A, et al. Feasibility ofMDCT Colonography in ultra-low-dose tech-nique in the detection of colorectal lesions:comparison with high-resolution video colono-scopy. AJR Am J Roentgenol 2004;183:1355–9.

[88] Edwards JT, Mendelson RM, Fritschi L, et al. Co-lorectal neoplasia screening with CT colonogra-phy in average-risk asymptomatic subjects:community-based study. Radiology 2004;230:459–64.

[89] Gluecker TM, Johnson CD, Harmsen WS, et al.Colorectal cancer screening with CT colonogra-phy, colonoscopy, and double-contrast bariumenema examination: prospective assessment ofpatient perceptions and preferences. Radiology2003;227:378–84.

[90] Pickhardt PJ, Taylor AJ, Johnson GL, et al.Building a CT colonography program: necessaryingredients for reimbursement and clinical suc-cess. Radiology 2005;235:17–20.

[91] Practice Guidelines ACR, Colonography CT. 10/01/05, p. 295–9. Available at: http://www.acr.org/s_acr/bin.asp?TrackID5&SID51&DID514802&CID51848&VID52&DOC5File.PDF.

[92] Zalis ME, Barish MA, Choi JR, et al. WorkingGroup on Virtual Colonoscopy. CT colonogra-phy reporting and data system: a consensusproposal. Radiology 2005;236:3–9.

[93] Macari M, Bini EJ, Jacobs SL, et al. Filling de-fects at CT colonography: pseudo- and

diminutive lesions (the good), polyps (thebad), flat lesions, masses, and carcinomas (theugly). Radiographics 2003;23:1073–91.

[94] Morrin MM, Farrell RJ, Kruskal JB, et al. Utility ofintravenously administered contrast material atCT colonography. Radiology 2000;217:765–71.

[95] Lefere P, Gryspeerdt S, Baekelandt M, et al.Laxative-free CT colonography. AJR Am J Roent-genol 2004;183:945–8.

[96] Iannaccone R, Laghi A, Catalano C, et al. Com-puted tomographic colonography without ca-thartic preparation for the detection ofcolorectal polyps. Gastroenterology 2004;127:1300–11.

[97] Luboldt W, Bauerfeind P, Wildermuth S, et al.Colonic masses: detection with MR colonogra-phy. Radiology 2000;216:383–8.

[98] Luboldt W, Bauerfeind P, Wildermuth S, et al.Contrast optimization for assessment of co-lonic wall and lumen in MR colonography. JMagn Reson Imaging 1999;9:745–50.

[99] Ajaj W, Pelster G, Vogt F, et al. Dark lumen MRcolonography: comparison to conventional co-lonoscopy for the detection of colorectal pa-thology. Gut 2003;52:1738–43.

[100] Albrecht T, Hohmann J, Oldenburg A, et al. De-tection and characterization of liver metastases.Eur Radiol 2004;14(Suppl 8):P25–33.

[101] Jang HJ, Kim TK, Wilson SR. Imaging of malig-nant liver masses: characterization and detec-tion. Ultrasound Q 2006;22:19–29.

[102] Bhattacharjya S, Bhattacharjya T, Baber S, et al.Prospective study of contrast-enhanced com-puted tomography, computed tomographyduring arterioportography, and magnetic reso-nance imaging for staging colorectal liver me-tastases for liver resection. Br J Surg 2004;91:1361–9.

[103] Mahfouz A-E, Hamm B, Taupitz M, et al. Hyper-vascular liver lesions: differentiation of focalnodular hyperplasia from malignant tumorswith dynamic gadolinium-enhanced MR imag-ing. Radiology 1993;186:133–8.

[104] Mahfouz A-E, Hamm B, Wolf K-J. Peripheralwash-out: a sign of malignancy on dynamic ga-dolinium-enhanced MR images of focal liver le-sions. Radiology 1994;190:49–52.

[105] Titu LV, Breen DJ, Nicholson AA, et al. Is rou-tine magnetic resonance imaging justified forthe early detection of resectable liver metastasesfrom colorectal cancer? Dis Colon Rectum2006;49:810–5.

[106] Warburg O. On the origin of cancer cells. Sci-ence 1956;123:309–14.

[107] Weber WA, Ott K, Becker K, et al. Prediction ofresponse to preoperative chemotherapy in ade-nocarcinomas of the esophagogastric junctionby metabolic imaging. J Clin Oncol 2001;19:3058–65.

[108] Wieder HA, Brucher BL, Zimmermann F, et al.Time course of tumor metabolic activity duringchemoradiotherapy of esophageal squamous

http://www.acr.org/s_acr/bin.asp?TrackID=&SID=1&DID=14802&CID=1848&VID=2&DOC=File.PDF



Gollub et al118

cell carcinoma and response to treatment. JClin Oncol 2004;22:900–8.

[109] Ott K, Fink U, Becker K, et al. Prediction of re-sponse to preoperative chemotherapy in gastriccarcinoma by metabolic imaging: results ofa prospective trial. J Clin Oncol 2003;21:4604–10.

[110] Choi MY, Lee KM, Chung JK, et al. Correlationbetween serum CEA level and metabolic vol-ume as determined by FDG PET in postopera-tive patients with recurrent colorectal cancer.Ann Nucl Med 2005;19:123–9.

[111] Zervos EE, Badgwell BD, Burak WE Jr, et al. Flu-orodeoxyglucose positron emission tomogra-phy as an adjunct to carcinoembryonic antigenin the management of patients with presumedrecurrent colorectal cancer and nondiagnosticradiologic workup. Surgery 2001;130:636–43[discussion: 643–4].

[112] Kemeny NE, Gonen M. Hepatic arterial infusionafter liver resection. N Engl J Med 2005;352:734–5.

[113] Fong Y, Saldinger PF, Akhurst T, et al. Utility of18F-FDG positron emission tomography scan-ning on selection of patients for resection of he-patic colorectal metastases. Am J Surg 1999;178:282–7.

[114] Selzner M, Hany TF, Wildbrett P, et al. Does thenovel PET/CT imaging modality impact on thetreatment of patients with metastatic colorectalcancer of the liver? Ann Surg 2004;240:1027–34.

[115] Akhurst TJ, Gonen M, Tuorto S, et al. Responseto neoadjuvant chemotherapy as measured byPET predicts outcome after liver resection forcolorectal metastases. J Clin Oncol 2006;24:3056.

[116] Hixson LJ, Fennerty MB, Sampliner RE, et al.Prospective blinded trial of the colonoscopicmiss-rate of large colorectal polyps. GastrointestEndosc 1991;37:125–7.

[117] Van Dam J. Novel methods of enhanced endo-scopic imaging. Gut 2003;52(Suppl 4):12–6.

[118] Quirke P, Durdey P, Dixon MF, et al. Local re-currence of rectal adenocarcinoma due to inad-equate surgical resection: histopathologicalstudy of lateral tumor spread and surgical exci-sion. Lancet 1986;2:996–9.

[119] Herbst F. Pelvic radiological imaging: a surgeon’sperspective. Eur J Radiol 2003;47:135–41.

[120] Salerno G, Daniels IR, Fisher SE, et al. A com-parison of digital rectal examination versusmagnetic resonance imaging in the predictionof circumferential resection margin status in pa-tients [abstract P172]. Colorectal Disease 2005;7(Suppl 2):56.

[121] Schaffzin DM, Wong DW. Endorectal ultra-sound in the preoperative evaluation of rectalcancer. Clin Colorectal Cancer 2004;4:124–32.

[122] Kwok H, Bissett IP, Hill GI. Preoperative stagingof rectal cancer. Int J Colorectal Dis 2000;15:9–20.

[123] Bipat S, Glas AS, Slors FJ, et al. Rectal cancer: lo-cal staging and assessment of lymph node in-volvement with endoluminal US, CT and MRimaging: a meta-analysis. Radiology 2004;232:773–83.

[124] Kulinna C, Eibel R, Matzek W, et al. Staging ofrectal cancer: diagnostic potential of multipla-nar reconstructions with MDCT. AJR Am JRoentgenol 2004;183:421–7.

[125] Brown G, Richards CJ, Bourne MW, et al. Mor-phologic predictors of lymph node status inrectal cancer with use of high-spatial-resolutionMR imaging with histopathologic comparison.Radiology 2003;227:371–7.

[126] Brown G, Daniels IR. Preoperative staging ofrectal cancer: The MERCURY Research Project.Recent Results Cancer Res 2005;165:58–74.

[127] Shoup M, Guillem JG, Alektiar KM, et al. Pre-dictors of survival in recurrent rectal cancer af-ter resection and intraoperative radiotherapy.Dis Colon Rectum 2002;45:585–92.

[128] Beets-Tan RG, Beets GL, Borstlap AC, et al. Pre-operative assessment of local tumor extent inadvanced rectal cancer: CT or high-resolutionMRI? Abdom Imaging 2000;25:533–41.

[129] Gearhart SL, Frassica D, Rosen R, et al. Improvedstaging with pretreatment positron emissiontomography/computed tomography in low rec-tal cancer. Ann Surg Oncol 2006;13:397–404.

[130] Guillem JG, Puig-La Calle J Jr, Akhurst T, et al.Prospective assessment of primary rectal cancerresponse to preoperative radiation and chemo-therapy using 18-fluorodeoxyglucose positronemission tomography. Dis Colon Rectum2000;43:18–24.

[131] Guillem JG, Moore HG, Akhurst T, et al. Se-quential preoperative fluorodeoxyglucose-posi-tron emission tomography assessment ofresponse to preoperative chemoradiation:a means for determining long term outcomesof rectal cancer. J Am Coll Surg 2004;199:1–7.

[132] Mortazavi A, Shaukat A, Othman E, et al. Post-operative computed tomography scan surveil-lance for patients with stage II and IIIcolorectal cancer; worthy of further study? AmJ Clin Oncol 2005;28:30–5.

[133] Pfister DG, Benson AB III, Somerfield MR. Clin-ical practice: surveillance strategies after cura-tive treatment of colorectal cancer. N Engl JMed 2004;350:2375–82.

[134] Rodriguez-Moranta F, Salo J, Arcusa A, et al.Postoperative surveillance in patients with co-lorectal cancer who have undergone curativeresection: a prospective, multicenter, random-ized, controlled trial. J Clin Oncol 2006;24:385–93.

[135] Gollub MJ, Akhurst T, Markowitz AJ, et al.Combined CT colonography and 18FDG PETimaging of colon polyps: a potential techniquefor selective cancer and pre-cancerous lesiondetection. Am J Roentgenol 2007;188:1–9.

05 - radiol clin n am 2007 - update on colorectal cancer imaging

Documents