1811

JOURNAL OF NEUROTRAUMA 24:1811–1816 (December 2007)© Mary Ann Liebert, Inc.DOI: 10.1089/neu.2007.0382

Comparative Magnetic Resonance Imaging at 1.5 and 3 Tesla for the Evaluation of Traumatic Microbleeds

RAINER SCHEID,1,2 DEREK V. OTT,2 HENRIK ROTH,3MATTHIAS L. SCHROETER,1,2 and D. YVES VON CRAMON1,2

ABSTRACT

Traumatic microbleeds (TMBs) can be regarded as a radiological marker of diffuse axonal injury(DAI). We sought to investigate the impact of the field strengths on the depiction of TMBs by T2*-weighted gradient echo magnetic resonance imaging (MRI). By the use of comparative MRI of 14patients (age range, 22–62 years) on 1.5- and a 3 T (Tesla) systems at a median time interval of 61months after traumatic brain injury (TBI), we found 239 (range 0.5–48.5, median 7.5) TMBs at 1.5T, and 470 (range 2–118, median 18.5) TMBs at 3 T, respectively (p � 0.001). However, in all butone patients MRI at 1.5 T also clearly showed TMBs. A significant negative correlation between thenumber of TMBs and the time interval TBI–MRI was observed, which was weaker for the imagingat 3 T (rs � �0.798; p � 0.001; and rs � �0.649; p � 0.012, respectively). In conclusion, T2*-weighted gradient-echo MRI at 3 T is superior as compared to MRI at 1.5 T for the detection ofTMBs. Nevertheless, in clinical practice, MRI at 1.5 T seems to be sufficient for this purpose. MRIat 3 T may be appropriate if there is a strong clinical suspicion of DAI, despite unremarkable rou-tine MRI, and possibly also if evidence of DAI is sought after a long interval from trauma.

Key words: diffuse axonal injury; diagnosis; high-field MR imaging; traumatic microbleeds

1Day Clinic of Cognitive Neurology, University of Leipzig, Leipzig, Germany.2Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany.3Department of Radiology, University of Leipzig, Leipzig, Germany.

INTRODUCTION

CEREBRAL MICROBLEEDS may be present in variouscentral nervous system (CNS) disorders (Fiehler,

2006). Since neuropathological studies have shown thattraumatic diffuse axonal injury (DAI) is typically ac-companied by small hemorrhages (“tissue tear hemor-rhages”) (Adams, 1989; Graham and Genarelli, 2000),traumatic microbleeds (TMBs) can be regarded as a neu-roimaging marker of this specific type of traumatic braininjury (TBI) (Scheid et al., 2003; Tong et al., 2004). Inthe absence of easily available and reliable radiological

tools for a direct proof of DAI, the demonstration ofTMBs is of substantial diagnostic importance.

T2*-weighted gradient echo magnetic resonance imag-ing (MRI) sequences are known to be sensitive for thedetection of cerebral microbleeds (Fiehler, 2006). Para-magnetic blood breakdown products (deoxyhemoglobin,methemoglobin, hemosiderin) induce local magneticfield inhomogeneities that manifest as focal signal loss(Fazekas et al., 1999). The T2* signal intensity loss de-pends on the magnetic field strength (Atlas et al., 1988).However, there is no information whether this basic phys-ical correlation is also of clinical relevance, and the hy-

SCHEID ET AL.

1812

FIG. 1. Comparison of magnetic resonance (MR) imaging at1.5 and 3 Tesla for the depiction of traumatic microbleeds(TMBs), median number of TMBs; n � 14 (Wilcoxon-Test).

pothetical threshold from which on hemosiderin or otherblood degradation products are detectable by MRI is notknown.

In a previous study, it was shown that T2*-weightedgradient echo imaging at high field strength (3 Tesla) isa useful tool for the evaluation of DAI in the chronicstage of TBI (Scheid et al., 2003). We now hypothesizedthat MRI at 3 T would indeed be superior for this pur-pose as compared to 1.5 T. In addition, we were inter-ested in the temporal dynamics of TMBs. The resultsshould have clinical implications for the radiological di-agnosis of DAI, at least when the diagnosis is sought inretrospect (e.g., for medicolegal reasons.)

METHODS

Fourteen patients (3 women, 11 men) with a history ofTBI and neuroimaging evidence of isolated TMBs fromprevious MRI at 3 T were prospectively studied betweenApril and July 2006. All patients had been victims of roadtraffic accidents. Patients’ ages ranged from 22 to 62years (median, 28 years). Demographic and clinical dataare summarized in Table 1.

After written informed consent by patients or relatives(n � 1), all patients underwent MRI on 1.5 T and on 3 T whole body systems (Siemens Symphony/SiemensMagnetom Trio; Erlangen, Germany). In order to controlfor bias from the potential temporal dynamic behavior ofTMBs, imaging was performed the same day on both sys-tems in each individual patient. The median time inter-val between TBI and MRI was 61 months (range, 15–116months).

Since all patients had undergone previous MRI withthe use of T1-, T2-, and FLAIR images in order to ruleout concomitant otherwise traumatic or nontraumaticpathologies, the current imaging protocol consisted solelyof 1.5 T and 3 T 2D T2*-weighted gradient echo images,which were performed in the same geometry on both sys-tems (24 slices, axial plane following AC-PC orientation,slice thickness 4 mm, slice gap 1 mm). The imaging pa-rameters were the same as for the use of the systems in

daily clinical practice (FOV 22.9 � 22.9 cm, data matrix256 � 256, TR 570 msec, TE 18.4 msec, flip angle 20°;FOV 19.2 � 19.2 cm, data matrix 256 � 256, TR 700msec, TE 15 msec, flip angle 25°).

Any 1–15-mm hypointense focus without connectionto the brain surface and/or the ventricular system was de-fined as TMB (Scheid et al., 2003), and their total num-ber was counted in each individual. Overlap with vas-cular structures was avoided. Areas of symmetrichypointensity of the globus pallidus, likely to representidiopathic iron and calcium deposits, were disregarded.With respect to other established causes of cerebral mi-crobleeds (Fiehler, 2006), it should be noticed that no patient had a history of arterial hypertension, diabetesmellitus, stroke, chronic alcohol abuse, or continuous anticoagulant or antiplatelet therapy.

Statistical Analysis

All images were evaluated independently by two read-ers (D.O., R.S.), and the mean number of TMBs out oftwo readings was used for nonparametric statisticalanalysis (Wilcoxon matched-pairs signed rank test,Spearman rank correlation). Due to obvious differences

TABLE 1. CLINICAL AND DEMOGRAPHIC DATA

Sex F/M (n � 14) 3/11Age (median/range), years (n � 14) 28/22–62Interval between MRI I and MRI II (median/range), months (n � 9) 66/44–73GCS (median/range) (n � 12) 3.5/3–6GOS (median/range) (n � 14) 5/4–7

GCS refers to the reported Glasgow Coma Scale score from initial hospital admission(Teasdale and Jennett, 1974). For two patients, this piece of information was not available.GOS refers to the current extended Glasgow Outcome Scale score (Wilson et al., 1998).

T2* 1.5 T T2* 3 T

20.0 p < .001 (Wilcoxon-Test)

18.0

6.0

0.0

2.0

4.0

14.0

16.0

12.0

8.0

10.0

FIG. 2. Illustrative comparison of 1.5 (A) and 3 Tesla (B) images of one patient. Note the higher sensitivity for traumatic mi-crobleeds (TMBs) of the 3 Tesla images in several anatomic locations.

SCHEID ET AL.

1814

FIG. 3. An axial view section on the level of the basal forebrain demonstrates the increased susceptibility to artifacts from airof sinuses and temporal bone of the 3 Tesla images (B) as compared to the 1.5 Tesla images (A).

in the appearance between the 1.5-T and 3-T images,complete blinding was not considered realistic.

RESULTS

Comparative MR-imaging revealed a total of 239(0.5–48.5, median 7.5) TMBs at 1.5 T, and 470 (2–118,median 18.5) TMBs at 3 T, respectively (Fig. 1). Inter-rater agreement for every participant was tested by theuse of the rank correlation � of Kendall, which showedsignificant agreement both for the imaging at 1.5 T (� �0.775; p � 0,01), and for the imaging at 3 T (� � 0,784;p � 0,01), respectively. There was a significant correla-tion between the individual number of TMBs on the 1.5-T and 3 T images (rs � 0.84; p � 0.000). Imaging eval-uation on an individual patient’s basis showed that MRIat 1.5 T did not clearly depict TMBs in only one patient(explaining the value of 0.5 for the lower range of TMBsfor the imaging at 1.5 T due to disagreement between theraters), whereas MRI at 3 T showed isolated TMBs inthis patient, too. There was a significant negative corre-lation between the number of TMBs and the time inter-val TBI–MRI for the imaging at 1.5 T (rs � �0.798; p �0.001), and a weaker significance for this interrelationfor the imaging at 3 T (rs � �0.649; p � 0.012), re-spectively. Because in nine patients MRI had alreadybeen performed previously with a 3-T system at a muchshorter median time interval TBI–MRI of 14 months(range 5–48 months) (Scheid et al., 2003, 2006), thesedata were additionally compared with the current results.There was no difference in the number of TMBs detectedin the two scans at different time points (rs � 0.95 p �0.000).

DISCUSSION

In a subset of head trauma patients, conventional MRsequences may fail to depict the associated traumaticbrain damage. These patients probably suffer from pureDAI. The detection of lesions suspicious of DAI is im-portant. Their presence in the acute stage of TBI is re-lated to outcome (Paterakis et al., 2000; Schaefer et al.,2004), and associated cognitive sequelae have been de-scribed for the chronic stage (Wallesch et al., 2001;Scheid et al., 2006). Medicolegal aspects must also beconsidered. MR technologies like diffusion-weightedimaging (DWI), diffusion tensor imaging (DTI), andMR spectroscopy (MRS) are valuable tools in order tomake such a diagnosis (Schaefer et al., 2004; Hos-houser et al., 2005; Xu et al., 2007). However, DTI isnot yet widely available, and DWI and MRS lack speci-ficity, particularly in the chronic stage after TBI. PureDAI may also be diagnosed via the detection of TMBs(Scheid et al., 2003), and in one study especially theoccurrence of hemorrhagic DAI in the acute phase ofTBI was related to outcome (Paterakis et al., 2000).Our study was conducted to clarify the impact of themagnetic field strengths on the depiction of TMBs byT2*-weighted gradient echo MRI. In order to avoid po-tential bias from other traumatic imaging abnormali-ties, only patients with neuroimaging findings of iso-lated TMBs who were thus regarded having sufferedpure DAI were studied. Although we were thereforeonly able to assess a small number of patients, the ap-proximately twofold increase in the total number ofTMBs on the 3-T images clearly shows that the theo-retical superiority of T2*-weighted gradient echo MRIat high field strengths is also of practical significance(Fig. 2).

However, since in all but one patients TMBs wereclearly detectable already on the 1.5 T images, it seemsthat despite the heightened sensitivity of 3 T MRI thevast majority of cases of probable DAI can be diagnosedby routine MRI at 1.5 T. For clinical and diagnosticpurposes it is important to establish this diagnosis per se in the first place. For the time being, the clinicalsignificance of the TMB load and its spatial distribu-tion is not yet clear (Tong et al., 2004; Scheid et al.,2006).

The finding of the one patient in whom 1.5 T MRI wasambiguous is further weakened by inaccuracies withinthe imaging procedures. Although attention was paid thatthe images were performed in the same geometry on bothsystems, scanning at slightly different locations or anglescould not be completely excluded. For example, singleand small TMBs could have been masked by anatomi-cally different slice gaps.

Moreover, the increased susceptibility at 3 T may notonly be advantageous. First, due to artifacts caused byair from sinuses and mastoid bone, in comparison espe-cially the 3 T gradient echo images are not suitable forthe evaluation of fronto-basal and temporo-basal/tem-poro-polar brain structures (Fig. 3). Second, the increasedsensitivity of 3 T MRI applies also to potential non-trau-matic microbleeds, possibly resulting in differential di-agnostic uncertainties (Fiehler, 2006).

We found a negative correlation between the numberof TMBs and the time interval TBI–MRI, particularly forthe imaging at 1.5 T. However, in a previous 3 T studywith more patients such a relationship was absent (Scheidet al., 2003). The respective negative result of the cur-rent subgroup analysis with the patients in whom 3 TMRI had been performed twice at markedly different timeintervals (median 14 and 61 months after TBI, respec-tively) are confirmative to the latter, and can not be ac-counted to potential differences in the TBI severity (me-dian Glasgow Coma Scale [GCS] scores in both groups3.5). It is likely that the conflicting results must be at-tributed to the small sample size of the current study. Ad-ditional bias from differences in the imaging geometrybetween the studies also notwithstanding, the resultswould nevertheless fit to the observed heightened sensi-tivity of T2*-weighted gradient echo imaging at 3 T. Thetime interval from trauma to MRI may be of general rel-evance and the visibility of DAI related TMBs on 1.5 TMRI could be more time dependent. The literature on thissubject is scarce (Wardlaw and Stratham, 2000). How-ever, two studies have reported matching results for MRIat 1.0 and 1.5 T, respectively (Messori et al., 2003; Ezakiet al., 2006). Prospective longitudinal studies with morepatients at different field strengths are needed for clari-fication.

Our study has several limitations: a limited number ofpatients, use of a single MR sequence, and use of retro-spective data for the subgroup analysis. Nevertheless, itmay be justified to draw the following conclusions: (i)Despite the heightened sensitivity of 3 T MRI, T2*-weighted gradient-echo MRI at 1.5 T seems to be suffi-ciently suitable for the detection of TMBs and for the di-agnosis of probable DAI after TBI. (ii) If the exactdistribution and localization of TMBs are of particularinterest, as may be the case with respect to medical ex-pert opinions and scientific investigations, imaging at 3T could be appropriate. (iii) MRI at high-field strengthsmight also be recommended in rare cases where clinicaldata and injury mechanisms suggest a diagnosis of DAI,despite normal findings on routine MRI.

ACKNOWLEDGMENTS

We thank all patients for participation.

REFERENCES

Adams, J.H., Doyle, D., Ford, I., et al. (1989). Diffuse axonalinjury in head injury: definition, diagnosis and grading.Histopathology 15, 49–59.

Atlas, S.W., Mark, A.S., Grossman, R.I., and Gomori, J.M.(1988). Intracranial haemorrhage: gradient-echo MR imag-ing at 1.5 T. Comparison with spin echo imaging and clini-cal applications. Radiology 168, 803–807.

Ezaki, Y., Tsutsumi, K., Morikawa, M., and Nagata, I. (2006).Lesions identified on T2*-weighted gradient echo images intwo patients with suspected diffuse axonal injury that re-solved in less than ten days. Acta Neurochir. (Wien.) 148,574–550.

Fazekas, F., Kleinert, R., Roob, G., et al. (1999). Histopatho-logic analysis of foci of signal loss on gradient-echo T2*-weighted MR images in patients with spontaneous intrac-erebral hemorrhage: evidence of microangiopathy-relatedmicrobleeds. AJNR Am. J. Neuroradiol. 20, 637–642.

Fiehler, J. (2006). Cerebral microbleeds: old leaks and newhaemorrhages. Int. J. Stroke 1, 122–130.

Graham, D.I., and Genarelli, T.A. (2000). Pathology of braindamage after head injury, in: Head Injury, 4th ed. P.R.Cooper, and J.G. Golfinos (eds), McGraw-Hill: New York,pps. 133–153.

Holshouser, B.A., Tong, K.A., and Ashwal, S. (2005). ProtonMR spectroscopic imaging depicts diffuse axonal injury inchildren with traumatic brain injury. AJNR Am. J. Neurora-diol. 26, 1276–1285.

Messori, A., Polonara, G., Mabiglia, C., and Salvolini, U.(2003). Is haemosiderin visible indefinitely on gradient-echo

MRI AT 1.5 AND 3 T FOR EVALUATION OF TRAUMATIC MICROBLEEDS

1815

SCHEID ET AL.

1816

MRI following traumatic intracerebral haemorrhage? Neuro-radiology 45, 881–886.

Paterakis, K., Karantanas, A.H., Komnos, A., and Volikas, Z.(2000). Outcome of patients with diffuse axonal injury: thesignificance of prognostic value of MRI in the acute phase.J. Trauma 49, 1071–1075.

Schaefer, P.W., Huisman, T.A.G.M., Sorensen, A.G., Gonza-lez, R.G., and Schwamm, L.H. (2004). Diffusion-weightedMR imaging in closed head injury: high correlation with ini-tial Glasgow Coma Scale score and Modified Rankin Scaleat discharge. Radiology 233, 58–66.

Scheid, R., Preul, C., Gruber, O., et al. (2003). Diffuse axonalinjury associated with chronic traumatic brain injury: evi-dence from T2*-weighted gradient-echo imaging at 3 T.AJNR Am. J. Neuroradiol. 24, 1049–1056.

Scheid, R., Walther, K., Guthke, T., et al. (2006). Cognitive se-quelae of diffuse axonal injury. Arch. Neurol. 63, 418–424.

Teasdale, G., and Jennett, B. (1974). Assessment of coma and impaired consciousness. A practical scale. Lancet 2,81–84.

Tong, K.A., Aswal, S., Holshouser, B.A., et al. (2004). Diffuseaxonal injury in children: clinical correlations with hemor-rhagic lesions. Ann. Neurol. 56, 36–50.

Wallesch, C.W., Curio, N., Kutz, S., Jost, S., Bartels, C., andSynovitz, H. (2001). Outcome after mild-to-moderate blunthead injury: effects of focal lesions and diffuse axonal in-jury. Brain Inj. 15, 401–412.

Wardlaw, J.M., and Statham, P.F. (2000). How often ishaemosiderin not visible on routine MRI following traumaticintracerebral haemorrhage? Neuroradiology 42, 81–84.

Wilson, J.T., Pettigrew, L.E., and Teasdale G.M. (1998). Struc-tured interviews for the Glasgow Outcome Scale and the Ex-tended Glasgow Outcome Scale: guidelines for their use. J.Neurotrauma 15, 573–585.

Xu, J., Rasmussen, I.A., Lagopoulos, J., and H Berg, A. (2007).Diffuse axonal injury in severe traumatic brain injury visu-alized using high-resolution diffusion tensor imaging. J. Neu-rotrauma 24, 753–765.

Address reprint requests to:Rainer Scheid, M.D.

Day Clinic of Cognitive NeurologyUniversity of Leipzig

Liebigstr. 22a04103 Leipzig, Germany

E-mail: scheid@cbs.mpg.de