Modified cine inversion recovery pulse sequence for the quantification of myocardial T1 and gadolinium partition coefficient

Original Research

Modified Cine Inversion Recovery Pulse Sequencefor the Quantification of Myocardial T1 andGadolinium Partition Coefficient

Matteo Milanesi, PhD,1* Andrea Barison, MD,1,2 Vincenzo Positano, MS,1

Pier Giorgio Masci, MD,1 Daniele De Marchi, RT,1 Luca Marinelli, PhD,3

Christopher J. Hardy, PhD,3 Thomas K. Foo, PhD,3 Luigi Landini, MS,1,4

and Massimo Lombardi, MD1

Purpose: To optimize and validate a modified cine inver-sion recovery sequence (MCine-IR) for myocardial T1quantification and gadolinium partition coefficient (lGd)estimation at 1.5 T.

Materials and Methods: The original version of the cineinversion recovery sequence was modified to allow fullytransverse magnetization recovery between two successiveinversion pulses. Sixty heart phases were acquired from anumber of heart cycles determined on a patient heart ratebasis. Phantom studies were carried out to find the opti-mal effective TR for myocardial and blood pool T1 quanti-fications in pre- and postcontrast studies. Four patientswith myocardial infarct (MI) and 22 dilated cardiomyopa-thy (DCM) were investigated, as well as 11 healthy sub-jects used as controls.

Results: Effective TR was identified to be 5000 msec and2000 msec, respectively, for pre- and postcontrast stud-ies. A longer precontrast (948 6 102 msec) and shorterpostcontrast (348 6 27 msec) T1 in ischemic patients rel-ative to DCM (815 6 98 msec, P ¼ 0.03 and 409 6 42msec, P ¼ 0.001) were noted in delayed enhancement(DE) areas. In MI patients lGd resulted higher than inDCM in DE areas (609 6 167 vs. 422 6 52, P ¼ 0.01) butlower in segments not exhibiting DE (355 6 100 vs. 3986 54, P ¼ 0.02).

Conclusion: It was feasible to measure T1 and lGd withMCine-IR and the results were in good agreement withthe literature.

Key Words: T1 quantification; gadolinium partition coeffi-cient; cine inversion recovery; myocardiumJ. Magn. Reson. Imaging 2012;000:000–000.VC 2012 Wiley Periodicals, Inc.

CARDIAC MAGNETIC RESONANCE IMAGING (MRI)enables the assessment of structural changes occur-ring in heart diseases by distinguishing normal frompathological myocardium. With this aim, the develop-ment of techniques for the quantification of myocar-dial T1 relaxation time has received particularattention in the last few years.

The majority of previous studies evaluated the clini-cal value of myocardial delayed enhancement (DE) forpostcontrast imaging. However, this technique relieson the relative difference in signal intensity betweenthe pathological and normal myocardium withoutallowing a direct quantification of T1 values of the tis-sues. Additionally, the postcontrast myocardial DEtechnique identifies gross fibrotic scars but not dif-fuse interstitial myocardial fibrosis. Thus, particularattention has recently been devoted to techniques thatdirectly measure T1 values of the myocardium beforeand after contrast agent administration. In particular,postcontrast T1 quantification has been successfullyused to differentiate between viable and necrotic tis-sue in patients with myocardial infarction (MI) (1–4).

Nonetheless, the knowledge of both pre- and postcon-trast values becomes beneficial for two main reasons.First, it helps to provide information on the presence ofinjured myocardium (1,2). Second, along with blood pre-and postcontrast T1 values, it can be used to quantifythe distribution volume of contrast agent in the myocar-dium, as measured by the gadolinium partition coeffi-cient (5–8), and consequently to estimate the extracellu-lar volume fraction after correction for the hematocrit (8).For this assessment, an accurate determination of bothpre- and postcontrast myocardial T1 values is crucial.

Pulse sequences such as the Modified Look–LockerInversion Recovery (MOLLI) (1,2,9–11) and the cine

1MRI Laboratory, Fondazione ‘‘G. Monasterio’’ CNR—Regione Toscana,Pisa, Italy.2Scuola Superiore Sant’Anna, Pisa, Italy.3General Electric Global Research, Niskayuna, New York, USA.4Department of Information Engineering, University of Pisa, Pisa, Italy.

Additional Supporting Information may be found in the online versionof this article.

*Address reprint requests to: M.M., Fondazione G Monasterio – CNRRegione Toscana, MRI Laboratory, via G Moruzzi 1, 56124 Pisa, Italy.E-mail: [email protected]

Received May 31, 2011; Accepted August 7, 2012.

DOI 10.1002/jmri.23807View this article online at wileyonlinelibrary.com.

JOURNAL OF MAGNETIC RESONANCE IMAGING 000:000–000 (2012)

CME

VC 2012 Wiley Periodicals, Inc. 1

inversion recovery approach (Cine-IR) (3,12–14) haveproved to be extremely reliable in the estimation ofmyocardial T1 values after contrast agent administra-tion. In particular, Cine-IR has been applied to bothT1 quantification (3,14) and the characterization ofmyocardial inversion times (12,13) in postcontrastimaging studies.

However, the estimation of precontrast myocardialand blood T1 values remains challenging becausesuch values are relatively long with respect to theheart cycle. Techniques that correct estimated T1 val-ues for the heart rate have been introduced, but theyare all specific to the pulse sequence to which theyapply (14) and/or based on empirical data (2,11,15).

In this work we modified the original Cine-IR pulsesequence to achieve a better estimation of precontrastT1 values without any further correction for patient’sheart rate in the fitting model. The key feature of themodified Cine-IR (MCine-IR) is to allow complete relax-ation of the longitudinal magnetization between twoconsecutive inversion pulses. In a conventional Cine-IR pulse sequence, an inversion recovery magnetiza-tion pulse is applied at the beginning of each cardiacR-R interval. Subsequent to playing out that radiofre-quency (RF) pulse, a fast gradient echo, segmentedk-space CINE acquisition is performed, producing aset of multiphase images, each with a different inver-sion time (TI). Consequently, the magnetization is notfully recovered prior to the application of the nextinversion recovery RF pulse, affecting the computationof tissue T1 relaxation times.

To ensure that a full recovery of the magnetization isalways achieved, in MCine-IR the number of heart-beats between the inversion pulses is determinedaccording to the heart rate of the subject, to enable aminimum magnetization recovery time. This adjust-ment makes the pulse sequence independent of thespecific heart rate and results comparable and consist-ent through subjects. Using this approach, the overallscan time for the MCine-IR sequence is approximatelythe same, regardless of the patient’s heart rate.

In this article phantom studies are presented to vali-date the methodology and to empirically determine thecorrect range of repetition time for both pre- and post-contrast T1 estimation. Subsequently, measurementsof pre- and postcontrast T1 of the myocardium aredescribed as a clinical application in cardiac patients,and a computation of partition coefficient is provided.

MATERIALS AND METHODS

All acquisitions were carried out on 1.5 T scanner(SIGNA Excite, GE Healthcare, Waukesha, WI) using astandard birdcage transmit coil and an 8-channelcardiac phased-array receiver coil. Statistical analysiswas performed with SPSS software for Windows (14.0rel.; Chicago, IL).

Pulse Sequence

Like the original Cine-IR (12–14), MCine-IR consists ofa nonselective adiabatic inversion pulse applied im-

mediately after the ECG R-wave trigger. The inversionpulse is followed by a fast spoiled gradient echo(FSPGR) CINE (segmented k-space) acquisition whereeach cardiac phase experiences a different time delayafter the inversion pulse and thus a different T1weighting. However, while in the original implementa-tions the CINE acquisition was limited to the firstheart cycle following the inversion and either the sec-ond cycle was left for T1 relaxation (12,14) or theinversion pulse and CINE acquisition were repeatedright from the subsequent R-wave (13), in MCine-IRthe acquisition is extended to further heart cyclesallowing for more complete recovery of longitudinalmagnetization. This allows even precontrast myocar-dium and blood to fully recover their longitudinalmagnetization which is interrogated by the CINEacquisition. Linear retrospective interpolation wasperformed on the CINE data to ensure up to 60 car-diac phases were reconstructed. Acquisition was per-formed in a single breath-hold. The number of R-to-R(R-R) intervals during which the same k-space linesare acquired in a segmented CINE fashion was setaccording to the following formula:

NRR ¼ðTReff ÞDRR

������ if modðTReff ;DRRÞ ¼ 0

ðTReff ÞDRR

������þ 1 if modðTReff ;DRRÞ 6¼ 0

8<: ½1�

where | | denotes integer division and mod( ) theremainder of the integer division. DRR is the R-R inter-val of the subject in msec and TReff is the effectiveTR between two successive inversion pulses. Theaddition of þ1 in the formula avoids any underestima-tion of TReff by rounding up its value to the nextheartbeat. The true effective TR of MCine-IR is clearlygiven by NRR DRR. During TReff data were continu-ously sampled to maintain the steady state conditionthroughout the whole experiment and to provide fur-ther points from the longitudinal magnetization (Mz)recovery curve.

Equation [1] has two main implications. First, itensures that the error committed in estimating the T1is not greater than the one committed by TReff. Sec-ond, it keeps the scan time similar between differentpatients and so the estimation of T1 less as depend-ent of the heart rate as possible.

The MCine-IR principle is illustrated in Fig. 1 wherecurves for different T1s (ie, 500–1000–3000 msec)were produced using the Bloch relaxation equation forMz (for simplicity’s sake, CINE readout RF pulseseffects were not considered) and a TReff equal to 5000msec. Such a TReff is enough to guarantee full recov-ery for T1 values of 500 msec and 1000 msec (respec-tively curves a and b) but not for a T1 of 3000 msec(curve c). The steady state condition also takes signifi-cantly longer to establish for a T1 of 3000 msec withrespect to shorter T1s.

Phantom Studies

Twelve distilled water tubes doped with different gado-linium concentrations were studied. The tubes’ tem-perature was fixed to 22C� among experiments.

2 Milanesi et al.

Reference T1 values were determined with a stand-ard spin-echo inversion recovery (SE) sequence withan effective TR about 5–6 times the maximum T1 asexpected by the gadolinium concentration used. TheT1s of the phantom involved were chosen to cover therange exhibited by pre- and postcontrast myocardiumand blood. Parameters used for the IR-SE sequencewere: TI ¼ 50-150-500-400-1200-1800-2400-3500msec, TR ¼ 10sþTI, matrix ¼ 128 � 128.

For MCine-IR, the pulse sequence was triggeredusing an ECG-emulator device set with a heart rate of60 bpm (DRR ¼ 1000 msec). The number of heartcycles (#RR) used in the acquisition was progressivelyincreased from 2 to 7 to simulate different TReff sce-narios between 2000 msec up to 7000 msec. The aimof this experiment was to determine the minimumTReff that ensured an accurate measurement of the T1in the range of in vivo values. The TReff value chosenfrom phantom experiments was then used in Eq. [1]to calculate the NRR for in vivo acquisitions, accordingto the specific patient’s heart rate.

For MCine-IR acquisition the following parameterswere used: flip angle (FA) ¼ 8�, matrix ¼ 128 � 128,TE/TR ¼ 1.1/3 msec, views per segment (VPS) ¼ 12.

Patient Studies

Eleven healthy subjects (age 47 6 10 years, 8 males),22 paucisymptomatic patients with dilated cardiomy-opathy (DCM, age 57 6 17 years, 14 males, all inNYHA class I-II), and four subjects with a recent MI(age 60 6 3 years, 3 males) were enrolled in the studyafter providing written informed consent.

The diagnosis of DCM was based on the evidence ofleft ventricle (LV) systolic dysfunction (LV ejectionfraction <50%) and the absence of obstructive coro-nary artery disease at invasive angiography, history ofmyocardial infarction, or revascularization. The diag-nosis of MI was made on typical chest pain and elec-

trocardiographic changes associated with increasedplasmatic level of cardiac troponin-I. All patients withMI were successfully treated by primary coronaryangioplasty within 12 hours from symptoms onset,and were studied by cardiac magnetic resonanceimaging (MRI) within 2 months from the acute event.The protocol complied with the 1975 Declaration ofHelsinki and was approved by our Institutional EthicsCommittee.

After localization of the four-chamber view with aCINE balanced-steady state free precession pulsesequence (FIESTA), the MCine-IR sequence wasacquired in all patients in a single short-axis mid-ven-tricular view before and at fixed time intervals (5 min,10 min, 15 min) after the administration of gadolin-ium contrast agents (Gadodiamide-OMNISCAN, 0.2mmol/kg). The following parameters were used forMCine-IR acquisition: field of view (FOV) ¼ 38 � 38 /48 � 38 cm2 depending on patient size, matrix ¼ 224� 160, FA ¼ 8�, TE/TR ¼ 2.4–2.6/5–6 msec, numberof averages (NEX) ¼ 0.5 with 16 overscans. VPS wasvaried from 24 to 12 between pre- and postcontrastacquisitions.

The presence of DE was assessed by standard seg-mented T1-weighted inversion-recovery fast gradientecho sequence, performed between 10 and 20minutes after contrast agent injection. Eight to twelveshort-axis views were acquired, encompassing theentire LV. The following parameters were used: slicethickness (no slice gap) ¼ 8 mm, TE/TR ¼ 1.6/4.6msec, VPS ¼ 24–32, FA 20�, matrix ¼ 224 � 192, andinversion time optimized to nullify normal myocar-dium (range 250–300 msec).

Image Analysis

Phantom and patient images were analyzed usingcustom-written software (HIPPO SW) developed in IDL7.1 (16). In phantom images analysis, a single region

Figure 1. MCine-IR pulse sequence. A CINE acquisition scheme is carried out between two consecutive nonselective adiabatic180� inversion pulses. The number of heartbeats between the two inversion pulses is determined according to Eq. [1] by thechosen TReff and patient heart rate. The number of inversion pulses, ie, the number of MCine-IR module repetitions, is con-trolled through the phase encoding steps and the views per segments as for a standard cine sequence. Examples of longitudi-nal magnetization recovery curves are depicted respectively for (a) T1 ¼ 500 msec, (b) T1 ¼ 1000 msec, and (c) T1 ¼ 3000msec. Curves were drawn using Bloch relaxation equation for Mz (for simplicity’s sake, CINE readout RF pulses effects werenot considered) and TReff was set to 5000 msec. The ECG trace underneath assumes a hypothetical heart rate of 60 bpm.

Modified Cine Inversion Recovery Sequence 3

of interest (ROI) was drawn at the center of a tubeand propagated along images with different TI times.In patient data, endocardial and epicardial contourswere manually defined on each frame together with areference point (ie, the superior intersection of theright and left ventricular walls). For each time frame,the myocardium was automatically segmented intosix equi-angular segments, following the AmericanHeart Association (AHA) guidelines (17). An ROI wasalso automatically defined in the center of the bloodpool to exclude papillary muscles. Signal versus TIcurves for myocardial and blood pool ROIs were eval-uated by averaging the signal in each ROI for each TItime. Although 60 cardiac phases were made availablefor postprocessing, only 20–30 phases were seg-mented and used for the inversion recovery fitting, toreduce the time for the analysis while maintaining asatisfactory number of data points.

CMR-DE images were analyzed offline using dedi-cated software (MASS 6.1, Medis, Leiden, Nether-lands) by an experienced operator who was blinded tothe clinical and T1-mapping data. The presence of DEwas visually determined on postcontrast images (0 ¼absent, 1 ¼ present) and localized according to AHAguidelines (17).

Since the MCine-IR acquisition starts soon after theinversion pulse, a steady state is not yet establishedfor the first few images. These initial time frames areusually affected by artifacts and their values cannotbe accurately fitted by the curve. For this reason twopoints of the curves were always discarded prior toperforming curve fitting. Resulting curves were thenfitted to a three parameters monoexponential modelA-Bexp(-TI/T1). The effect produced by SPGR RFpulses on relaxation curves as described previously(18) was not taken into account in the fitting modelleaving such an investigation to future studies. Signalpolarity for magnitude images was assigned choosingthe inversion point that minimizes the fitting error(19).

In phantom studies, a percentage error of T1 estima-tion with MCine-IR was computed using the valuesdetermined with the IR-SE sequence as reference. Thedata were then divided into two groups identified by therange of pre- and postcontrast T1 values, and a meanpercentage error was computed separately for the twogroups. Since ‘‘true’’ references were not available fromin vivo data, the range of T1 values was drawn basedon literature data. In previous in vivo studies, in fact(2,4,6,7,10,11,20,21), a range of 700–1600 msec wasfound for precontrast blood and myocardial T1s,whereas in the first 20 minutes following an 0.1–0.2mmol/kg injection of Gd-DTPA, the same regions liebetween 150–550 msec. Such a distinction allowsdetermining two suitable minimum TReff values for pre-and postcontrast conditions in subsequent in vivostudies. The adopted criterion was to choose the mini-mum TReff, which ensured a mean percentage errorsimilar in pre- and postcontrast imaging. Such a TReff

was not necessarily the same for the two conditions.In vivo T1 values were used for the estimation of the

gadolinium partition coefficient in the myocardium(lGd), defined as the ratio between the change in

relaxation after gadolinium administration in the tis-sue and in the blood at equilibrium. Assuming therelaxivity rate R of the gadolinium is the same in myo-cardium and blood (22), the following formula can beused to compute the partition coefficient (6,7):

lGd ¼ DR1 myocardium

DR1 blood

¼ R1 myocardium�postcontrast � R1 myocardium�precontrast

R1 blood�postcontrast � R1 blood�precontrast½2�

where R1 ¼ 1/T1 and the values for the blood pool areevaluated in the LV cavity. To investigate the kineticsof gadolinium distribution, three lGd values were cal-culated considering myocardial and blood values atthree different times after contrast injection (5, 10, 15min) and steady state equilibrium between the plasmaand myocardial interstitium was assumed to bereached when gadolinium partition coefficientremained constant over time.

Statistical Analysis

In healthy subjects, a one-way analysis of variance(ANOVA) test was performed between T1 values esti-mated on different cardiac segments. This test was runseparately for precontrast, 5-minute, 10-minute, 15-minute measurements to assess whether susceptibilityand geometrical artifacts could affect T1 evaluation. Inhealthy subjects, to exclude possible dependency ofmyocardial T1 estimation on heart rate, Pearson’s cor-relation coefficient was used to test the correlationbetween estimated T1 values and heart rate.

To study the effect of the presence and type of car-diac disease on our measurements, an ANOVA testwith Bonferroni’s post-hoc analysis was performed inhealthy subjects, DCM and MI patients for both T1values and lGd values. Cardiac segments without DEin DCM (nDE-DCM) and MI (nDE-MI) patients werealso compared to those of healthy subjects. Patientssegments with myocardial DE (DE-DCM and DE-MI)were analyzed separately to assess differences in T1values and lGd.

Finally, blood T1 values from healthy subjects,DCM and MI patients were compared to exclude aninterference of the results by the infusion rate of gad-olinium and to normalize gadolinium tissue concen-trations to the kinetics of gadolinium administration.All tests were two-tailed at the 5% significance level.

RESULTS

Phantom Results

T1 values of the 12 distilled water tubes doped withgadolinium estimated by IR-SE ranged between 208and 1780 msec. Figure 2 shows the comparisonbetween phantom T1 as estimated indicated by IR-SEand MCine-IR as a function of different TReff. Therange of values used for the calculation of the per-centage error is also shown in the graph. Overall, onlytwo values were dropped for the subsequent analysis(732 and 1780 msec) because they did not fit in either

4 Milanesi et al.

the pre- or postcontrast range. At first glance, Fig. 2reveals that MCine-IR is precise in the estimation ofT1 in the postcontrast range irrespective of the TReff.On the contrary, in the precontrast range the TReff

based sequence underestimates the T1 values as aresult of longitudinal magnetization saturation. Onlythe use TReff values greater than 4000–5000 msecyielded good agreement with the data from the IR-SEacquisitions, although with some differences betweenlower and higher T1 values.

In Fig. 3 the percentage errors (single and averagevalues) are reported as a function of TReff and sepa-rately for the pre- and postcontrast ranges of T1 values.In the postcontrast range, illustrated in the upperpanel, the percentage error remains well under 2%regardless of TReff. For the precontrast range, the lowerpanel indicates that for a TReff of 5000 msec and higherthe mean error lies under 2%. Unlike the postcontrastscenario, in the precontrast range a further reductionof the error was achieved at increasing TReff, but ahigher TReff would mean a higher acquisition time (ie,breath-hold) for the patient involved in the study.

As a consequence, for postcontrast imaging, sincefor all tested TReff the error was found to be below 2%without substantial difference, 2000 msec was chosenfor in vivo applications. A TReff of 5000 msec was cho-sen for precontrast acquisitions, being the first valuefor which the mean percentage error sets at the samevalue encountered in the postcontrast case, ie, 2%.

In Vivo Results

After an analysis of the CMR-DE images, myocardialDE was found in all four subjects with a recent MIcharacterized by a subendocardial to transmural

extension. Areas of myocardial DE with nonischemicpattern were also found in eight (36%) DCM patients,with a patchy or midwall distribution pattern. In noneof the healthy subjects was myocardial DE detected.The MCine-IR sequence was feasible in all patientswith an acquisition time that varied from about 20seconds to 16 seconds between pre- and postcontrastscans.

An example of in vivo images acquired with MCine-IR is showed in Fig. 4 for an MI patient with a trans-mural MI involving the inferior, inferoseptal, andinferolateral walls. Precontrast and 10-minute post-contrast images from different cardiac cycles after theinversion pulse are displayed and times after inver-sion reported. Examples of model fits to inversion re-covery data are also reported for the inferior (ischemicarea) and the anterior (remote area) segment, as wellas the blood pool. Enlarged precontrast imagesdepicted in Fig. 4 correspond respectively to the nul-ling point of the infarct area and to a further point ofthe flat part of the signal intensity curve. Enlargedpostcontrast images give a representation of the nul-ling point time of the DE remote area of the myocar-dium and, as in the precontrast case, of an inversiontime when the longitudinal magnetization has almostrecovered. In both cases the different signal levels inthe inferior and anterior segments were in accordancewith the level of the signal as shown in the graphs.

Our Supporting Materials display CINE movies froma DCM patient again in the precontrast condition(pre_contrast.mpg) and 10 minutes after contrastagent administration (post_contrast.mpg).

Figure 2. Phantom results. T1 values obtained from IR-SE(referred as to ‘‘true’’ T1 values) on a collection of Gd-DTPA-doped water tubes are compared with MCine-IR estimation.MCine-IR acquisition was performed using an ECG simulatorset at 60 bpm and the TReff was changed from 2000 msec upto 7000 msec. The shorter five T1 of the phantom (from 210msec to 548 msec) were considered representative of the post-contrast blood-myocardial range of values, whereas valuesbetween the seventh to the eleventh (from 702 to 1547 msec)were considered representative of the precontrast range.

Figure 3. Percentage errors committed by the MCine-IRsequence on phantom with respect to IR-SE as a function ofTReff. Upper panel: percentage error on postcontrast range ofT1 values. Lower panel: percentage error on precontrastrange of T1 values. Single T1 values errors are representedby asterisks, whereas average errors for TReff are linked toone another through a solid line.

Modified Cine Inversion Recovery Sequence 5

Segmental T1 values estimated on healthy subjectsare given in Table 1 (mean values and standard devia-tions). The ANOVA test performed between T1 valuesof healthy subjects demonstrated that there was nosignificant correlation between T1 values and mea-

surement location (P ¼ 0.30). Also, myocardial andblood T1 values estimated in healthy subjects did notshow any significant correlation with heart rate (P ¼0.99), confirming the reliability of Eq. [1] to choosethe number of heartbeats for a given TReff.

Figure 4. A series of MCine-IR images in a patient diagnosed with an inferior MI, extended to the inferoseptal and inferolat-eral left ventricular walls. In the upper panel some images are displayed from different cardiac cycles after the inversionpulse for the precontrast acquisition, and times after inversion reported. Two images have been chosen and enlarged, to rep-resent the mid-diastolic phase of the first and the fourth heartbeat of the acquisition (patient’s heart rate ¼ 62 bpm, numberof heartbeats acquired ¼ 5). Myocardial segmentation is overlaid on the two enlarged images. Signal intensities and the fittedcurves from the inferior segment (infarcted area), the anterior segment (remote area), and the blood pool are also given. As apostcontrast example (lower panel), some images from 10 minutes after injection acquisition were chosen. In this case, twomid-diastolic cardiac phases are shown enlarged, one in the first heartbeat and the other in the second heartbeat of the ac-quisition (patient’s heart rate ¼ 60 bpm, number of heartbeats acquired ¼ 2). The more left-handed enlarged images for bothpre- and postcontrast acquisitions have been chosen to show respectively the nulling point of the infracted (precontrastimage) and the remote (postcontrast image) areas, whereas the images on the right-hand side are from further inversiontimes, almost on the flat part of the inversion recovery curve.

6 Milanesi et al.

Figure 5 illustrates in vivo T1 values as estimatedby MCine-IR. The bars represent average values overall six myocardial segments of a mid-ventricular slicein healthy subjects, DCM and MI patients, as well asDCM and MI patients considering separately the seg-ments positive and negative to DE.

In the precontrast series, there was a significant dif-ference in T1 values between cardiac patients andhealthy subjects (P ¼ 0.03). In particular, healthy andDCM were found significantly lower than MI (819 6

56 and 810 6 79 msec vs. 855 6 60 msec, P ¼ 0.02).When analyzing subgroups of cardiac segments, DE-MI showed a remarkable increase in the T1 value,which again was higher than DE-DCM patients (9486 102 msec vs. 815 6 98 msec, P ¼ 0.03). No signifi-cant difference was found in the group of LV segmentswithout DE.

After contrast agent administration, T1 values inhealthy subjects were always higher than in MIpatients. Such a difference in values gained statisticalsignificance at 10 minutes (406 6 32 msec vs. 380 6

42 msec, P ¼ 0.004) and 15 minutes (422 6 30 msecvs. 403 6 54 msec, P ¼ 0.03) but not at 5 minutes af-ter contrast injection (373 6 29 msec vs. 356 6 36msec, P ¼ 0.09). Likewise, T1 in DCM was foundhigher than in MI patients, showing a statistical sig-nificance after 10 minutes (401 6 33 msec vs. 380 6

42 msec, P ¼ 0.005) and 15 minutes (423 6 23 msecvs. 403 6 54 msec P ¼ 0.01) but not after 5 minutes.Taking into consideration all segments of DCMpatients, no statistical difference arose between theirT1 values and those of healthy subjects. In segmentsnegative to DE, T1 values were not sensitive to the na-ture of cardiac disease, showing values very close tothose of healthy subjects. However, T1 in DE-MIswere lower than DE-DCMs and their differenceincreased as a function of time after injection, gainingstatistical significance at 15 minutes (348 6 27 msecvs. 409 6 42 msec, P ¼ 0.001).

T1 values obtained for blood are shown in Fig. 6.The bar graph reveals how all three categories of datayielded very similar results both in pre- and postcon-trast conditions. Accordingly, no statistical difference

Table 1

Segmental Mean T1 Values and Standard Deviations From Healthy Subjects

Segments

T1 pre T1 post 5min T1 post 10min T1 post 15min

Mean SD Mean SD Mean SD Mean SD

Anterior 823 89 370 38 413 36 423 36

Anteroseptal 859 61 372 30 415 44 421 36

Inferoseptal 834 48 365 23 414 29 416 36

Inferior 808 55 375 29 403 32 423 20

Inferolateral 797 80 376 28 396 25 414 13

Anterolateral 792 90 382 30 397 28 433 36

Data are in milliseconds.

Figure 5. Mean T1 relaxation times estimated in vivo byMCine-IR as a function of time after contrast agent injection(0 minutes is the precontrast condition). Error bars representthe standard deviation. In the graph are included T1 valuesderived from: all segments from healthy subject (Healthy), allsegments from dilated-cardiomyopathic subjects (DCM), allsegments from ischemic patients (MI), all segments negativeto DE from dilated-cardiomyopathic patients (nDE-DCM), allsegments negative to DE from ischemic patients (nDE-MI),all segments positive to DE from dilated-cardiomyopathicpatients (DE-DCM), all segments positive to DE from ische-mic patients (DE-MI).

Figure 6. Mean blood T1 relaxation times estimated in vivoby MCine-IR as a function of time after contrast agent injec-tion (0 minutes is the precontrast condition). Error bars rep-resent the standard deviation. In the graph are included T1

values derived from the blood of healthy subjects, DCM andMI patients.

Modified Cine Inversion Recovery Sequence 7

emerged between healthy subjects and patients withDCM and MI.

Figure 7 illustrates partition coefficient values asestimated by Eq. [2] in healthy subjects, DCM and MIpatients, as well as DCM and MI patients, consideringseparately the areas positive and negative to DE. Aswe can see, although lGd values are relatively close toone another at 5 minutes after injection, the graphbars shows the same distribution at 10 and 15minutes. However, considering lGd values at 10 and15 minutes after gadolinium injection, there was asignificant difference between cardiac patients andhealthy subjects (P ¼ 0.03). In particular, healthy andDCM were found significantly lower than MI at 10minutes (384 6 44 and 393 6 70 vs. 449 6 69, P ¼0.04) and at 15 minutes (373 6 47 and 400 6 54 vs.440 6 54, P ¼ 0.02). lGd values in DCM patients wereincreasingly greater than healthy subjects as a func-tion of time after injection, although no statistical sig-nificance was reached. Interestingly, at 15 minutes af-ter contrast injection in nonenhanced areas DCMpatients showed a lGd significantly higher (398 6 54)than that of healthy (373 6 47) and MI (355 6 100)patients (P ¼ 0.02). In cardiac segments with DE, par-tition coefficient in DCM patients was lower than MIsat 10 and 15 minutes (497 6 109 vs. 571 6 115, 4226 52 vs. 609 6 167, respectively), and statistical sig-nificance was reached at 15 minutes (P ¼ 0.01).

DISCUSSION

The MCine-IR sequence presented in this study wasdesigned with the flexibility to choose the most suitableTReffs that allows the relaxation of longitudinal mag-netization to the level required for more accurate T1estimation. This approach enables one to minimize theacquisition time (ie, patient breath-hold time) whilemaintaining the accuracy required for a reliable T1estimation in both pre- and postcontrast imaging.Other Cine-IR sequences, like the one introduced pre-viously (3,14), manage to keep TReff equal to 2 R-Rintervals by incorporating the incomplete magnetiza-tion recovery into the signal intensity calculation. Withthis approach a substantial reduction in scan time canbe achieved with respect of MCine-IR, although thecurve fitting model becomes more complex.

Although the idea of lengthening the effective repeti-tion time was recently presented also in the MOLLIsequence (4,23), the interval between two preparationpulses in that case is given by a fixed number ofheartbeats (6 and 5, respectively), resulting in abreath-hold time that varied with heart rate. Regard-ing the choice of sequence used to sample the longitu-dinal magnetization recovery, most MOLLI implemen-tations (2,10,11,23) make use of balanced SSFPreadouts, although a correction method for T1 estima-tion in a noncontinuous balanced SSFP approach hasnot been reported yet (24).

Because the number of heartbeats between twoinversion pulse is set according to the specificpatient’s heart rate with MCine-IR, no other correctionfor patient heart rate is required in the estimationmodel of T1. Phantom studies have in fact been usedto set our TReff to a mean percentage error of 2% forboth pre- and postcontrast cardiac studies and Eq. [1]ensures that the percentage error will be in the worstcase the one given for TReff. The effectiveness of Eq [1]for in vivo acquisition is confirmed by the lower panelof Fig. 4, where the signal intensity curves demon-strate how longitudinal magnetization is recovered inthe chosen TReff. Our in vivo analysis of myocardialand blood T1 values in healthy subjects proved theabsence of any significant correlation between T1 val-ues and heart rate.

Our reference myocardial T1 value in precontrast ac-quisition (819 6 56 msec) fits well into the range of val-ues indicated by the literature at 1.5 T. Within such arange, the lower and upper values were respectively720 6 110 msec (7) and 1033 6 126 msec (6), bothobtained using Look–Locker techniques. As for theblood T1 values, no difference emerged betweenhealthy subjects and patients and our reference pre-contrast value of 1598 6 85 msec is in accordancewith other results. For instance, in Ref. (22) a value of1580 6 140 msec was found by employing a 4-pointsingle-shot IR sequence, whereas the MOLLI sequencesuggested 1551 6 55 msec in Ref. (10). A bit shortervalue of 1379 6 127 msec was obtained in Ref. (7), stillwith a Look–Locker technique.

Postcontrast T1 values in general depend on thecontrast agent concentration and relaxivity as well ason the time of observation. Our postcontrast findings,

Figure 7. Mean Gd-DTPA partition coefficients estimated invivo by MCine-IR as a function of time after contrast agentinjection. Error bars represent the standard deviation. In thegraph are included partition coefficient derived from: all seg-ments from healthy subject (Healthy), all segments fromdilated-cardiomyopathic subjects (DCM), all segments fromischemic patients (MI), all segments negative to DE fromdilated-cardiomyopathic patients (nDE-DCM), all segmentsnegative to DE from ischemic patients (nDE-MI), all seg-ments positive to DE from dilated-cardiomyopathic patients(DE-DCM), all segments positive to DE from ischemicpatients (DE-MI).

8 Milanesi et al.

which range from 227 6 21 msec to 280 6 10 msecfor the blood and from 373 6 29 msec to 421 6 30msec for the myocardium, fall between the valuessuggested by the literature and used in the phantomstudies to determine the most suitable TReff, ie, 200–550 msec for a 0.1–0.2 mmol/kg injection of Gd-DTPA (4,6,7,10,11,21–23).

In precontrast acquisitions, patient data pointedout that T1 was significantly higher in ischemicpatients with respect to DCM and healthy subjects.This result is in agreement with Refs. (4,12) and islikely due to the presence of edema in acute infarctsegments.

Postcontrast values added valuable informationregarding the presence and type of cardiac diseasewith respect to precontrast data. Although no statis-tical significance was found at 5 minutes after con-trast injection, confirming that equilibrium had notyet been reached, at 10 and 15 minutes myocardialT1 values of MI patients became significantly lowerthan both healthy and DCM patients. More interest-ing is the fact that T1 values of MI patients in DEsegments reached a significantly lower value thanthat measured on DCM in the same type of seg-ments at 15 minutes, bringing to light an objectivelyquantifiable measure in the type of delayedenhancement.

The gadolinium partition coefficient lGd showedgreat value in confirming and completing T1 quantifi-cation information. Moreover, after adjustment forblood hematocrit, the extracellular volume fraction ofthe myocardium could be derived as an in vivo surro-gate of a histological analysis of myocardial interstitialremodeling. In the nonsegmental analysis lGd was sig-nificantly higher in MI patients with respect to healthysubjects, confirming observations in ischemic cardio-myopathy in previous studies (6,7). The partition coef-ficient was also higher in DCM than healthy subjects,confirming findings in Ref. (8) for idiopathic dilatedcardiomyopathy, although our data did not reach sta-tistical significance. When MI and DCM were com-pared, MI data reached significantly higher values oflGd than DCM both in nonsegmental and in DE seg-ment analysis.

A higher postcontrast T1 and a lower partition coef-ficient of DE areas in DCM patients with respect to MIcan be explained by considering the different patternof myocardial fibrosis between ischemic and nonis-chemic cardiomyopathy. In particular, the ischemicpatterns of DE are likely characterized by moreintense and localized replacement fibrosis as com-pared with more subtle and diffuse fibrosis of nonis-chemic patients. Our study indicates also that lGd

was lower in nDE-DCM than in nDE-IM segments andhealthy subjects. This aspect reflects once again asubtle myocardial fibrosis widespread throughout thewhole myocardium, corresponding both to small fociof replacement fibrosis and to interstitial myocardialfibrosis described at histology (24,25). Notably, thisresult suggests the partition coefficient may provide atool capable of highlighting the presence of a dilatedcardiomyopathy also in the cardiac segments negativeto DE studies.

A closer look at Fig. 7 reveals that nonsegmentaland DE segments data in cardiac patients shows anincreasing trend of lGd as a function of time afterinjection, whereas healthy subjects were characterizedby a rather constant value throughout the observationtime. This behavior confirms that in healthy subjectsthe partition coefficient reached a steady state condi-tion already after 5 minutes from contrast injection,whereas in cardiac patients this does not happenuntil 10–15 minutes (7,22). Even though someauthors employed a constant gadolinium infusion toreach equilibrium (26), recently it was confirmed thatequilibrium between blood and myocardial gadolin-ium concentrations can be achieved also after bolusinjection (27).

In this study, lGd values were calculated assumingno difference in the gadolinium relaxivity betweenblood and myocardium, even though in a previousstudy a correction for myocardial/blood relaxivity ra-tio was introduced, leading to 0.8 at 1.5 T and 0.93 at3 T (22). This correction deserves further studieswhen comparing lGd values from different studies anddifferent MR equipment, and in particular whenassessing the volume fraction of extravascular extrac-ellular matrix from lGd (8,27).

Our study has some limitations. Our statisticalresults on MI patients must be regarded cautiously,since the population is not numerous (24 myocardialsegments from four MI patients). In this perspective,although our results are in line with the literature,further studies are envisaged. Other limitations derivefrom the choice of exploiting the whole cardiac cycleto estimate the T1 in a segmented cine acquisitionstrategy. On the one hand, this approach increasesthe time resolution of relaxation curves sampling. Buton the other hand, it requires more processing timedue to the need to trace contours in each frame. Also,a pixel-by-pixel T1 mapping would not be easily avail-able as a consequence of the cardiac phase evolution,and through plane motion may limit the accuracy ofthe analysis, although no effect of disturbance on thesteady state was noticed in our images. For the samereason, an accurate delineation of DE area was notpossible in the study. However, segmental assessmentof T1 was fully feasible. The choice of acquiringimages in a CINE fashion could also cause infloweffects that eventually affect the contrast between themyocardium and the blood pool. Further investigationis needed to better characterize possible differences inthe inversion recovery curves of left and right ventricleblood pools. In our study the breath-hold timerequired for the acquisition was well tolerated by allthe subjects involved in the study, both in pre- andpostcontrast conditions. However, it is possible thatbreath-hold time may limit the application of the tech-nique in some patients. The use of parallel imagingapproaches could be used in the future to reducescan times.

In conclusion, the modified version of the Cine-IRsequence presented in this article proved capable ofproviding a wide range of myocardial and blood T1values, which agree well with those found in the liter-ature for healthy subjects. The estimated values did

Modified Cine Inversion Recovery Sequence 9

not show any heart rate dependency, thanks to thesequence flexibility in determining the number ofheartbeats needed for the recovery of longitudinalmagnetization. Postcontrast T1 quantification wasconfirmed to be a valuable tool in highlighting the dif-ferences between healthy subjects, ischemic anddilated cardiomyopathic patients in the area ofdelayed enhancement. Thanks to an accurate estima-tion of precontrast T1 range of values, as our phan-tom studies proved, an estimation of gadolinium par-tition coefficient was possible, which added valuableinformation to mere T1 quantification, especially incardiac segments negative to DE. Such an index,although originally developed to investigate myocar-dial infarction, is becoming more and more usefulin the early diagnosis of microscopic interstitial fibro-sis detection (25) and may prove useful for manag-ing many cardiac diseases such as dilatedcardiomyopathy.

REFERENCES

1. Sparrow PS, Messroghli DR, Reid S, Ridgway JP, Bainbridge G,Sivananthan MU. Myocardial TI mapping for detection of left ven-tricular myocardial fibrosis in chronic aortic regurgitation: pilotstudy. Am J Radiol 2006;187:630–635.

2. Messroghli DR, Walters K, Plein S, et al. Myocardial T1 mapping:application to patients with acute and chronic myocardial infarc-tion. Magn Reson Med 2007;58:34–40.

3. Goldfarb J W, Arnorl S, Marguerite R. Gadolinium Pharmacoki-netics of chronic myocardium infarcts: implications for late gado-linium-enhanced infarct imaging. J Magn Reson Imaging 2009;30:763–770.

4. Piechnik SK, Ferreira VM, Dall’Armellina R, et al. Shortenedmodified Look-Locker inversion recovery (ShMOLLI) for clinicalmyocardial T1-mapping at 1.5 and 3 T within 9 heartbeat breath-hold. J Card Magn Res 2010;12:69–80.

5. Harris PA, Lorenz CH, Holburn GE, Knowles AO. Regional mea-surement of the Gd-DTPA tissue partition coefficicent in caninemyocardium. Magn Reson Med 1997;38:541–545.

6. Flacke SJ, Fisher SE, Lorenz CH. Measurement of the gadopente-tate dimeglumine partition coefficient in human myocardium invivo: normal distribution and evaluation in acute and chronic in-farction. Radiology 2001;218:703–710.

7. Klein C, Nekolla SG, Balbach T, et al. The Influence of myocardialblood flow and volume of distribution on late Gd-DTPA kineticsin ischemic heart failure. J Magn Reson Imaging 2004;20:588–594.

8. Jerosch-Herold M, Sheridan DC, Kushner JD, et al. Cardiac mag-netic resonance imaging of myocardial contrast uptake and bloodflow in patients affected with idiopathic or familiar dilated car-dyomyopathy. Am J Physiol Heart Circ Physiol 2008;295:H1234–H1242.

9. Karlsson M, Nordell B. Phantom and in vivo study of the Look-Locker T1 mapping method. Magn Reson Imaging 1999;17:1481–1488.

10. Messroghli DR, Radjenovic A, Kozerke S, Higgins DM, Sivanan-than MU, Ridgway JP. Modified Look-Locker inversion recovery(MOLLI) for high-resolution T1 mapping of the heart. Magn ResonMed 2004;42:141–146.

11. Messroghli DR, Plein S, Higgins DM, Walters K, Jones TR, Siva-nanthan MU. Human myocardium: single breath-hold T1 map-ping with high spatial resolution—reproducibility study.Radiology 2006;238:1004–1012.

12. Gupta A, Lee VS, Chung Y-C, Babb JS, Simonetti OP. Myocardialinfarction: optimization of inversion times at delayed contrast-enhanced MR imaging. Radiology 2004;233:921–926.

13. Ho VB, Hood MN, Montequin M, Foo TK. Cine inversion recovery(IR): rapid tool for optimized myocardial delayed enhancementimaging. In: Proc 13th Annual Meeting ISMRM, Miami;2005:1675.

14. Goldfarb JW, Mathew ST, Reichek N. Quantitative breath-holdmonitoring of myocardial gadolinium enhancement using inver-sion recovery TrueFISP. Magn Reson Med 2005;53:367–371.

15. Wansapura J, Gottliebson W, Crotty E, Fleck R. Cyclic variationof T1 in the myocardium at 3T. Magn Reson Imaging 2006;24:889–893.

16. Positano V, Milanesi M, Masci P, et al. Feasibility of myocardialT1 mapping from Cine-IR images by image warping. In: Proc 16thAnnual Meeting ISMRM, Stockholm; 2010:3726.

17. Cerqueira MD, Weissman NJ, Dilsizian V, et al. Standardizedmyocardial segmentation and nomenclature for tomographicimaging of the heart: a statement for healthcare professionalsform the Cardiac Imaging Committee of the Council on ClinicalCardiology of the American Heart Association. Circulation 2002;105:539–542.

18. Deichmann R, Haase A. Quantification of T1 values by SNAP-SHOT-FLASH NMR Imaging. J Magn Reson 1992;96:608–612.

19. Nekolla S, Gneiting T, Syha J, Deichmann R, Haase A. T1 mapsby k-space reduced snapshot-FLASH MRI. J Comput AssistTomogr 1992;16:327–332.

20. Sharma P, Socolow J, Patel S, Pettigrew RI, Oshinki JN. Effect ofGd-DTPA-BMA on blood myocardial T1 at 1.5T and 3T inhumans. J Magn Reson Imaging 2006;23:323–330.

21. Illes L, Pfliger H, Phrommintikul A, et al. Evaluation of diffusemyocardial fibrosis in heart failure with cardiac magnetic reso-nance contrast-enhanced T1 mapping. J Am Coll Card 2008;52:1574–1800.

22. Donahue KM, Burstein D, Manning WJ, Gray ML. Studies of Gd-DTPA relaxivity and proton exchange rates in tissue. Magn ResonMed 1994;32:66–76.

23. Messroghli DR, Greiser A, Frohlich M, Dietz R, Schulz-Menger J.Optimization and validation of a fully-integrated pulse sequencesfor Modified Look-Locker inversion-recovery (MOLLI) T1 mappingof the heart. J Magn Reson Imaging 2007;26:1081–1086.

24. Song T, Stainsby JA, Ho VH, Hood MN, Slavin GS. Flexible car-diac T1 Mapping using a modified Look-Locker acquisition withsaturation recovery. Magn Reson Med 2012;67:622–627.

25. Weber KT, Brilla CG. Pathological hypertrophy and cardiac inter-stitium. Fibrosis and renin-angiotensin-aldosterone system. Cir-culation 1991;83:1849–1865.

26. Mewton N, Liu CY, Croisille P, Bluemke D, Lima JAC. Assessmentof myocardial fibrosis with cardiovascular magnetic resonance. JAm Coll Cardiol 2011;57:891–903.

27. Flett AS, Hayward MP, Ashworth MT, et al. Equilibrium contrastcardiovascular magnetic resonance for the measurement of dif-fuse myocardial fibrosis: preliminary validation in humans. Cir-culation 2010;122:138–144.

28. Schelbert EB, Testa SM, Meier CG, et al. Myocardial extravascu-lar extracellular volume fraction measurement by gadoliniumcardiovascular magnetic resonance in humans: slow infusion ver-sus bolus. J Cardiovasc Magn Reson 2011;4:13–16.

10 Milanesi et al.