Original Research

Myocardial Tagging and Strain Analysis at 3 Tesla:Comparison With 1.5 Tesla Imaging

V. Uma Valeti, MD,1 Wookjin Chun, MD,2 Donald D. Potter, MD,3 Philip A. Araoz, MD,4

Kiaran P. McGee, PhD,4 James F. Glockner, MD,4 and Timothy F. Christian, MD1*

Purpose: To determine whether imaging at 3 T could im-prove and prolong the tag contrast compared to imagesacquired at 1.5 T in normal volunteers, and whether suchimprovement would translate into the ability to performstrain measurements in diastole.

Materials and Methods: Normal volunteers (N � 13) werescanned at 1.5 T (GE Signa CV/i) and 3.0 T (GE VH/i). AnECG-triggered, segmented k-space, spoiled-gradient-echogrid-tagged sequence was used during cine acquisition. Tagcontrast was determined by the difference of the meansignal intensity (SI) of the tagline to the mean SI of themyocardium divided by the standard deviation (SD) of thenoise (CNRtag). Matched short-axis (SA) slices were ana-lyzed. Strain measurements were performed on images us-ing a 2D strain analysis software program (harmonic phase(HARP)).

Results: The average CNRtag over the cardiac cycle wassuperior at 3 T compared to 1.5 T for all slices (3 T: 23.4 �12.1, 1.5 T: 9.8 � 8.4; P � 0.0001). This difference re-mained significant at cycle initiation, end-systole, and theend R-R interval (at cycle termination: 3 T � 14.0 � 11.0 vs.1.5 T � 4.4 � 3.5; P � 0.01). Strain measures were obtain-able only in early systole for 1.5 T images, but were robustthroughout the entire R-R interval for 3 T images.

Conclusion: Imaging at 3 T had a significant benefit for myo-cardial tag persistence through the cardiac cycle. The improve-ment allowed strain analysis to be performed into diastole.

Key Words: magnetic resonance; ventricular function;myocardial strain; noninvasive imaging; high field imagingJ. Magn. Reson. Imaging 2006;23:477–480.© 2006 Wiley-Liss, Inc.

MYOCARDIAL TAGGING using MRI is an evolving non-invasive technique for assessing and quantifying re-gional myocardial function and strain in the humanheart. Since this technique was first described (1,2),several studies have used MRI tagging to measure re-gional myocardial function with multiphase myocardialstrain in normal and diseased states (3–6). Algorithmsthat are necessary to quantify strain and regional myo-cardial function are heavily dependent on the contrastbetween the tag lines and the myocardium. However,myocardial tagging techniques are limited by the rapidfading of tags, which restricts their application to thesystolic phases of the cardiac cycle. Given the increas-ing recognition of the role of diastolic dysfunction inseveral cardiac disease states (7), it would be of clinicalvalue to overcome the limitations posed by the earlydecay tags and extend the analysis of regional myocar-dial function and strain to the entire cardiac cycle.Several methods, including complementary spatiallymodulated magnetization (CSPAMM) (8,9) and bal-anced free precession sequences (10,11) have been em-ployed toward that end. Higher-field-strength imagingis becoming more common and provides another path-way to potentially improve tag persistence through pro-longed T1 myocardial recovery and enhanced signal-to-noise ratio (SNR).

The aim of this study was to compare tag contrast asa function of time within the cardiac cycle in normalvolunteers imaged at both 1.5 T and 3 T using a fastspoiled-gradient-recalled echo sequence with spatialmodulation of magnetization. The images were ana-lyzed for systolic and diastolic strain measurementsusing a software program that generates strain mapsfrom the spatial derivative of harmonic phase (HARP)images in k-space.

MATERIALS AND METHODS

Thirteen healthy volunteers (nine males and four fe-males, mean age � 33 � 8.4 years, range � 29–60years) were enrolled in the study and underwent scan-ning at 3 T and 1.5 T. The mean time between scans was42 days � 22 days (range � 2–65 days). All studies wereconducted after written informed consent was ob-tained, and were in accordance with human-subjectIRB guidelines at the Mayo Clinic. 3-Tesla scanning was

1Department of Medicine, Mayo Clinic and Foundation, Rochester,Minnesota, USA.2Section of Cardiology, University of Wisconsin, Madison, Wisconsin, USA.3Department of Surgery, Mayo Clinic and Foundation, Rochester, Min-nesota, USA.4Department of Radiology, Mayo Clinic and Foundation, Rochester,Minnesota, USA.*Address reprint requests to: T.F.C., University of Vermont, 191 Baird,MCHV, Burlington, VT 05495.E-mail: timothy.christian@uvm.eduReceived January 20, 2005; Accepted November 30, 2005.DOI 10.1002/jmri.20527Published online 28 February 2006 in Wiley InterScience (www.interscience.wiley.com).

JOURNAL OF MAGNETIC RESONANCE IMAGING 23:477–480 (2006)

© 2006 Wiley-Liss, Inc. 477

performed on a GE Signa 3.0 T VH/i, and 1.5 T scan-ning was performed on a GE Signa cvi system (GEMedical Systems, Milwaukee, WI, USA). Both scannershad the same gradient performance: gradients of 40mT/m with a slew rate of 150 mT/m/msec. Four chan-nel phased-array torso coils (two anterior and two pos-terior), tuned to 64 and 128 MHz, respectively, wereused for both scans. These coils had the same diameter,overlap, and number of elements for both fieldstrengths.

Imaging Protocol

The tagging technique was performed as previously de-scribed (12). The volunteers were studied in a supineposition using prospective ECG triggering. Scans wereobtained on separate days. The imaging parameterswere as follows: TR � 8 msec, TE � 4.0 msec, constantflip angle � 20° at 1.5 T and 10° at 3 T, slice thickness �8 mm, matrix � 256 � 160 pixels with phase field ofview (FOV) � 0.75. The FOV was 360 mm2. A bandwidthof 62.5 kHz was used for both field strengths. The gridor line spacing was 7 mm. Tissue tags were created inan orthogonal grid pattern (for short-axis (SA) imaging)of presaturation in the myocardium using an ECG-triggered, segmented k-space, spoiled-gradient-echogrid-tagged sequence during cine acquisition (GE FastCine) with spatial modulation of magnetization(SPAMM). Twenty phases were acquired across the car-diac cycle at eight views per segment over 16 heartbeatsfor a temporal resolution of 64 msec.

All images were obtained with the volunteers holdingtheir breath at end-inspiration. The end-systolic imagewas identified just prior to ventricular relaxation(smallest ventricular chamber) and the final image waschosen as the end R-R image. For all patients, the totaltime required for imaging was less than 40 minutes.

Image Analysis

Images were analyzed offline with dedicated software(Cine, GE Medical Systems, Milwaukee, WI, USA; IDL,Research Systems, Inc., Boulder, CO, USA). Three SAslices (apical, mid, and base) were analyzed at both fieldstrengths per volunteer from the five to six SA slicesacquired. These were registered using anatomic land-marks for each set of scans (images were chosen to be 1cm from the first appearance of the apical LV cavity and1 cm from the membranous septum, and at the mid-papillary muscle level whenever possible). For eachslice, regions of interest (ROIs) were selected in theanterior wall to calculate the mean signal intensity (SI).Tag ROIs were manually drawn for a single myocardialtag line in all phases of the cardiac cycle (approximately7 mm long). A square myocardial ROI was drawn withina single grid (approximately 5 � 5 mm) adjacent to thetag ROI, with care taken to avoid the surrounding tag-lines.

The mean SIs of the tags (SITag) and myocardium(SImyocardium) were tabulated to calculate parameters re-lating to tag contrast and persistence (11):

SNR � (SImyocardium)/SD noise (1)

Tag contrast (Ctag) � (SImyocardium � SITag) (2)

CNRtag � Ctag/SD noise (3)

where CNR � contrast-to-noise ratio, and SD noise �the SD of the noise of an ROI external to the chest wall.No surface coil intensity correction algorithms wereused. CNRtag of the images by field strength is presentedboth as a function of absolute time from the R-wavetrigger, and as a function of cardiac phase to add clin-ical perspective, since there was no significant changein heart rate between scans.

HARP Strain Analysis

The tagged images were processed using a HARP imagespatial derivative 2D-strain analysis package (HARP v1.01; Diagnosoft Inc., Baltimore, MD, USA). A midven-tricular SA slice was chosen for analysis. The SA slicewas divided into six radial transmural segments, two ofwhich (the anterior and inferior segments) were chosenfor analysis. Radial strain maps were constructedthrough the 20 phases of the tagged images (13) forboth field strengths.

Statistical Analysis

Data are expressed as the mean � SD. A comparison ofmatched variables between 1.5 T and 3 T scans wasperformed using a two-tailed paired t-test for overallcomparisons or for specific cardiac phases. A P-value �0.05 was considered significant. All analyses were per-formed using Statistica 6.0 software (Statsoft Inc.,Tulsa, OK, USA).

RESULTS

There were no adverse effects for any volunteer at eitherfield strength, and successful scans were acquired in all

Figure 1. Mean � SD CNRtag values for all 13 volunteers ineach of the 20 cardiac phases of the cine images acquired at1.5 T (gray circles) and 3 T (black squares). The mean HR foreach acquisition was 65 � 11 bpm, so the x-axis covers ap-proximately 1000 msec. There was a significant difference (P �0.0001) for field strength at any given cardiac phase.

478 Valeti et al.

subjects. There was no change in heart rate betweenexaminations (65 � 11 for both scans). Backgroundnoise for external ROIs was significantly reduced at 3 T(3.6 � 1.2 SI at 3 T vs. 7.4 � 2.3 SI at 1.5 T, P � 0.0001).The SNR of the nontagged myocardium was signifi-cantly higher at sequence initiation for 3 T images(21.8�7.0 at 1.5 T vs. 38.7�8.7 at 3 T, P � 0.0001), butfell slightly short of the theoretical doubling of signal atthe double field strength.

Tag Contrast Analysis

The mean CNRtag over the entire cardiac cycle for allvolunteers (Fig. 1) was superior at 3 T (23.4�12.6) com-pared to 1.5 T (9.8�8.4, P � 0.0001). CNRtag was supe-rior at 3.0T compared to 1.5 T at end-diastole (cycleinitiation), end-systole, and end-diastole (Fig. 2). Thiswas independent of location in relation to the cardiaccoils. Figure 3 shows a comparison of tagged imagesfrom a single volunteer through the cardiac cycle ateach field strength.

Strain Analysis

Radial strain measures were calculated for each phaseof the cardiac cycle for all volunteers (Fig. 4), and acurve was generated by intersecting the mean valuesfor each field strength. The curves are similar, as are

the SDs of the mean values for both field strengths,until early systole. As the cardiac cycle progresses, the1.5 T curve takes an erratic course and the SDs of thestrain measures of each phase become widely dispar-ate. The 3 T curve follows a normal diastolic relaxationpattern, and the SD of strain values remains narrow.

DISCUSSION

Strain analysis using tagged images can involve a largeamount of manual tracing to supplement semiauto-mated tag detection algorithms. A limitation inherent tomany tagging analyses is the rapid fading of tags, whichmakes it difficult to achieve automated detection of thetagging grid. In addition, many previously describedtagging techniques were restricted to analyzing only thesystolic phase of the cardiac cycle, because of this samelimitation. There is a need to improve and prolong tag–myocardium contrast to facilitate automated tag detec-tion algorithms and to extend strain analyses to dia-stolic assessment.

Imaging at 3 T offers two advantages that may affecttagged imaging using SPAMM: 1) the SNR is theoreti-cally doubled compared to 1.5 T imaging, and 2) T1

recovery is prolonged for myocardial tissue (myocardialT1 at 1.5 T � 880 msec, T1 at 3 T � 1115 msec) (14,15).The prolongation of T1 recovery in conjunction with

Figure 2. Mean � SD for CNRtag values at three distinct points in the cardiac examination (initiation of the sequence (R-waveof the ECG), end-systole, and end-diastole) for the anterior (left) and inferior (right) walls. CNRtag was consistently higher at 3 Tby location or cardiac phase.

Figure 3. Example of a tagged SA slice from tag deposition (phase 1) to cardiac cycle termination (phase 20) from a singlevolunteer acquired at 1.5 T (top row) and 3 T (bottom row). Note that the tags persist well into diastole at 3 T, whereas tag contrastis lost shortly after systole at 1.5 T.

Tagged Imaging at 3.0 T 479

improved SNR should improve tag contrast with nosignificant change in image acquisition duration.

The results for SPAMM tagging at 3 T using a spoiledFGRE sequence confirm these theoretical advantages,although the SNR at 3 T fell short of doubling comparedto 1.5 T. The loss in contrast occurred early in thecardiac cycle at 1.5 vs. 3 T. Subjectively, tag contrast at3 T remained visible at the end of the cardiac cycle,whereas virtually no contrast was evident at 1.5 Tshortly after end-systole. The persistence in tag con-trast appeared to have beneficially impacted the mea-surement of 2D strain in diastole when the commer-cially available strain analysis package was used. Theclinical impact of this remains to be determined, but itmay provide another tool for comprehensively describ-ing left-ventricular (LV) function by MRI.

CSPAMM and steady-state free precession (SSFP) im-aging sequences may improve tag contrast and persis-tence (8–11). Neither was available in the present studyfor comparison. The present analysis was confined tonormal volunteers, so it would be of interest to assess

the performance of 3 T tagging in patients with LVdysfunction.

The improvement in tag contrast and persistence us-ing SPAMM at 3 T was significant. Tag contrast wassuperior in all phases of the cardiac cycle compared to1.5 T images, and allowed the quantitation of myocar-dial strain into diastole. A comparison of diastolic tagcontrast at 1.5 T using SSFP- and CSPAMM-based se-quences will clearly be of interest for future studies.

REFERENCES1. Zerhouni EA, Parish DM, Rogers WJ, Yang A, Shapiro EP. Human

heart: tagging with MR imaging—a method for noninvasive assess-ment of myocardial motion. Radiology 1988;169:59–63.

2. Axel L, Dougherty L. MR imaging of motion with spatial modulationof magnetization. Radiology1989;171:841–845.

3. McVeigh ER, Atalar E. Cardiac tagging with breath-hold cine MRI.Magn Reson Med 1992;28:318–327.

4. Moore CC, O’Dell WG, McVeigh ER, Zerhouni EA. Calculation ofthree-dimensional left ventricular strains from biplanar tagged MRimages. J Magn Reson Imaging 1992;2:165–175.

5. Nagel E, Stuber M, Burkhard B, et al. Cardiac rotation and relax-ation in patients with aortic valve stenosis. Eur Heart J 2000;21:582–589.

6. Ennis DB, Epstein FH, Kellman P, Fanapazir L, McVeigh ER, AraiAE. Assessment of regional systolic and diastolic dysfunction infamilial hypertrophic cardiomyopathy using MR tagging. Magn Re-son Med 2003;50:638–642.

7. Zile MR, Brutsaert DL. New concepts in diastolic dysfunction anddiastolic heart failure. Part I: diagnosis, prognosis, and measure-ments of diastolic function. Circulation 2002;105:1387–1393.

8. Fischer SE, McKinnon GC, Scheidegger MB, Prins W, Meier D,Boesiger P. True myocardial motion tracking. Magn Reson Med1994;31:401–413.

9. Fischer SE, Mckinnon GC, Maier SE, Boesiger P. Improved myo-cardial tagging contrast. Magn Reson Med 1993;30:191–200.

10. Zwanenberg JJ, Kuijer JP, Marcus JT, Heethaar RM. Steady-statefree precession with myocardial tagging: CSPAMM in a singlebreathhold. Magn Reson Med 2003;49:722–730.

11. Herzka DA, Guttman MA, McVeigh ER. Myocardial tagging withSSFP. Magn Reson Med 2003;49:329–340.

12. Croisille P, Moore CC, Judd RM, et al. Differentiation of viable andnonviable myocardium by the use of three-dimensional tagged MRIin 2-day-old reperfused canine infarcts. Circulation 1999;99:284–291.

13. Osman NF, Kerwin WS, McVeigh ER, Prince JL. Cardiac motiontracking using harmonic phase (HARP) magnetic resonance imag-ing. Magn Reson Med 1999;42:1048–1060.

14. Wen H, Denison TJ, Singerman RW, Balaban RS. The intrinsicsignal-to-noise ratio in human cardiac imaging at 1.5, 3, and 4 T. JMagn Reson 1997;125:65–71.

15. Noeske R, Seifert F, Rhein KH, Rinneberg H. Human cardiac imag-ing at 3 T using phased array coils. Magn Reson Med 2000;44:978–982.

Figure 4. Radial strain measures from a single transmuralsegment of an SA midventricular slice averaged for each vol-unteer in each of 20 cardiac phases throughout the cardiaccycle at 1.5 T (gray) and 3 T (black). The curve at 3 T follows anormal strain pattern with consistent SD values, whereas thecurve at 1.5 T fluctuates wildly with wide SDs when the dia-stolic phase is encountered.

480 Valeti et al.