JOURNAL OF MAGNETIC RESONANCE IMAGING 32:878–886 (2010)
Original Research
Late Gadolinium Enhancement of Acute Myocardial Infarction in Mice at 7T: Cine-FLASH Versus Inversion Recovery Andrea Protti, PhD,1* Alexander Sirker, MD,1 Ajay M. Shah, MD, PhD,1 and Rene Botnar, PhD2 Purpose: To investigate myocardial infarction (MI), late gadolinium (Gd) enhancement (LGE), cardiovascular magnetic resonance imaging (CMRI) is used as the current gold standard for the in vivo diagnosis in patients and preclinical studies. While inversion recovery (IR) fast gradient echo LGE imaging is the preferred technique at clinical field strengths it remains to be investigated which is the best sequence at higher field strength. We therefore compared the IR technique against cine fast low shot angle (cine-FLASH) for the quantification of MI size in mice at 7T in vivo. Materials and Methods: Five mice were used to optimize cine-FLASH and IR parameters. Nine mice were subsequently imaged with optimized parameters using both techniques 2–3 days after MI and 30 minutes post Gd injection. Results: The difference in infarct size values was within 3.3% between the two CMRI techniques and within 7.5% of histological values by Bland–Altman analysis. Contrast-to-noise-ratio between infarcted and normal tissue as well as blood was higher for cine-FLASH with the additional benefit of a 2-time-fold shorter scan time than with the IR method. Furthermore, left ventricular function/volumes could be calculated from cine-FLASH images before as well as after Gd injection. Conclusion: In conclusion, cine-FLASH LGE MRI represents an attractive alternative to IR LGE MRI for infarct size assessment in mice at high field strengths because it provides similar accuracy while being more robust, faster, and less user dependent. Key Words: late gadolinium enhancement (LGE); myocardial infarction (MI); cine-FLASH; inversion recovery (IR); LV mass; mouse/murine J. Magn. Reson. Imaging 2010;32:878–886. C 2010 Wiley-Liss, Inc. V
1
Cardiovascular Division, King’s College London British Heart Foundation Centre, London, United Kingdom. 2 Division of Imaging Sciences, King’s College London, London, United Kingdom. Contract grant sponsor: British Heart Foundation; Contract grant numbers: RE/08/003, CI/05/003 and CH/99001. *Address reprint requests to: A.P., the Rayne Institute, 4th Floor, Lambeth Wing St Thomas Hospital, London SE1 7EH, UK. E-mail:
[email protected] Received September 26, 2009; Accepted July 9, 2010. DOI 10.1002/jmri.22325 View this article online at wileyonlinelibrary.com. C 2010 Wiley-Liss, Inc. V
THE RELIABLE DIAGNOSIS and quantification of myocardial infarction (MI) is essential in preclinical research studies that address the pathophysiology of myocardial ischemia and its consequences. Infarct size is a key determinant of both acute left ventricular (LV) dysfunction and longer-term adverse remodeling (1). Methods that allow noninvasive assessment of infarct size in studies involving genetically modified mice are particularly desirable since this allows serial longitudinal evaluation and correlation between MI size and subsequent LV remodeling (2). Late gadolinium (Gd) enhancement (LGE) cardiovascular magnetic resonance imaging (CMRI) (3) is a common way to investigate MI and left ventricular (LV) remodeling after coronary occlusion in humans (4) and has recently also been applied in mice (2,5,6). In the acute setting of MI, myocyte membranes rupture and subsequently the intracellular as well as the extracellular space expand locally and allow the accumulation of Gd (7). At such a stage, LGE imaging can be accomplished by exploiting the different longitudinal magnetic relaxation times (T1-relaxation) between viable tissue with little Gd accumulation due to intact cell membranes, and nonviable tissue where Gd is retained, resulting in shortened T1 values. The most common LGE CMRI technique is a T1-weighted gradient echo sequence employing an inversion recovery (IR) prepulse (8). This method is well established in clinical practice and has also been used in preclinical studies in mice and rats (9). In mice rapid heart rates between 500–600 bpm often lead to suboptimal electrocardiogram (ECG) triggering, which may impact the performance of IR sequences, as they rely on a constant time delay between successive IR prepulses. Such challenges may be especially prominent in mice with acute MI with severe loss of weight and physical strength. Several groups proposed solutions by either modifying the standard IR approach (9,10) or by using cine fast low shot angle (cine-FLASH) methods (2,11–13). Generally, the cine-FLASH method is utilized for functional/volumetric measurements before Gd injection (2,13,14), followed by an IR technique used for LGE and tissue differentiation. Cine-FLASH imaging without preparatory prepulses is currently not
878
LGE of Acute MI
879
Figure 1. A: Cine-FLASH diagram. The heart cycle was resolved using 9 frames with one k-space line per frame. To maintain the steady state, RF pulses were continuously applied, while data acquisition was only enabled during end expiration. RF pulses of 40 were applied and the TE was 1 msec. B: IR turbo-FLASH diagram. A nonadiabatic 180 pulse was applied at the first heartbeat immediately after detection of the end-expiratory phase. Subsequently, a TI of typically 300–400 msec was used to null signal from normal myocardium and imaging was performed with a turbo-FLASH technique, which was initiated by the third or fourth heartbeat to collect four k-space lines per acquisition.
considered an optimal method for tissue characterization using LGE acquisitions due its inability to accurately delineate infarcted from normal viable myocardium, in particular at clinical field strengths (3,15). At high field instead, several studies (12,13) that have assessed infarct size with cine-FLASH reported reduced MI delineation compared to IR methods. However, cineFLASH provides images with high spatial and temporal resolution and highly accurate measurements of small physiological changes of cardiac function and structure. In this study we sought to optimize both IR and cine-FLASH sequence parameters using simulation and experimental data to investigate the optimal LGE imaging sequence at 7T for assessment of MI in mice. We subsequently compared the two methods, and propose cine-FLASH as a more robust, faster, and easier way for infarct size assessment with LGE at high field strengths. We also show that cine-FLASH LGE can be used to assess functional and volumetric measurements achieving similar results to precontrast data, thereby drastically reducing the total scan time of a preclinical cardiac MR experiment.
MATERIALS AND METHODS Animal Model All in vivo procedures were conducted in accordance with the Guidance on the Operation of the Animals. Female C57Bl/6J mice weighing 18–24 g were anesthetized with 2% isoflurane and 98% oxygen. Animals underwent endotracheal intubation and were ventilated using a dedicated small animal ventilator (Hugo Sachs Elektronik, Germany). A lateral thoracotomy was made, chest wall muscles were incised and reflected, and the thorax opened in the fourth intercostal space. The pericardium was removed to access the epicardial surface. The left coronary artery was ligated using 8/0 Ethilon suture, at a level between 1 and 2 mm below the tip of the left atrium. Successful ligation was confirmed by regional blanching of the left ventricle, extending to the apex. The chest wall was then repaired in layers and the animals weaned from the ventilator. Animals recovered in a warmed chamber for at least 6 hours. Perioperative analgesia
with Buprenorphine intramuscularly and Flunixin subcutaneously was used. Animal CMRI Two days after MI, CMRI was performed on a 7T horizontal MR scanner (Varian, Palo Alto, CA) with mice in prone position. The gradient coil had an inner diameter of 12 cm, gradient strength was 1000 mT/m (100G/cm), and rise-time 120 ms. A quadrature transmit/receive coil (RAPID Biomedical, Germany) with an internal diameter of 39 mm was used. Anesthesia was maintained with 1.5% isoflurane / 98.5% oxygen and body temperature was maintained at 37 C using a warm air fan (SA Instruments, Stony Brook, NY). The ECG was monitored via two metallic needles placed subcutaneously in the front paws. A pressure-transducer for respiratory gating was placed on the animals abdomen. To synchronize data acquisition with the ECG and to compensate for respiratory motion, simultaneous ECG triggering and respiration gating (SA Instruments) were applied. LGE CMR imaging was performed after intraperitoneal (IP) injection of a 30 mL bolus of 0.5 mmol/kg Gd-DPTA (Magnevist, Schering Healthcare, UK). Five animals were used for the cine-FLASH flip angle optimization experiment with imaging performed 20 to 30 minutes after IP injection of the Gd compound. A Look–Locker sequence was performed in the same mice between 30 to 50 minutes to determine the inversion time to null viable tissue and to derive T1 values. Typical heart rates were between 375– 500 bpm (cycle length 120 to 160 msec) with a fluctuation of 639 bpm (10 msec) per cardiac cycle and 40 6 10 resp/m (1500 6 500 msec) per respiratory cycle. Another nine animals were subsequently used for comparison between the optimized IR and cineFLASH sequences. CMR Sequences Cine-FLASH Cine-FLASH was used to acquire temporally resolved dynamic short-axis images of the heart. The gradient echo technique (Fig. 1A) maintains steady-state
880
during the entire scan. Spoiler gradients of 1 msec duration and 100 mT/m strength were applied after each data acquisition readout to dephase the remaining transverse magnetization before the application of the next radiofrequency (RF) excitation pulse. Cine-FLASH was performed before, with simultaneous ECG triggering and respiration gating (double gating), and after, with ECG triggering only (single gating), Gd injection. Imaging parameters of the optimized sequence included TR ¼ RR-interval/number of frames (typically 14 msec), TReff ¼ RR-interval, TE ¼ 1 msec, FOV ¼ 20 25 mm, matrix size ¼ 128 128, slice thickness ¼ 1 mm; flip angle ¼ 40 , 2 averages for the double gating, 3 averages for the single gating, 9 slices, 1 k-space line/frame, 9 frames per cardiac cycle to study the dynamic contraction of the heart. The acquisition time was 8 6 0.5 minutes for both cine-FLASH before and after Gd injection. Preand postacquisitions used different parameters, number of averages, and different gating in order to investigate cine-FLASH robustness on motion artifacts when using a similar scanning time. The trigger was positioned at the peak of the QRS complex. The phase encoding steps were equally distributed along the cardiac cycle in order to obtain diastolic and systolic frames for measurement of functional/volumetric information. Ejection fraction (EF), stroke volume (SV), cardiac output (CO), infarct size, and LV mass were estimated. During the optimization phase all cine-FLASH scanning parameters were maintained except for the number of slices and flip angle. A single short axis slice comprising both infarcted and healthy myocardium was examined. Sequence flip angles varied from 10 to 70 in order to study signal-to-noise ratio (SNR) and relative contrast-to-noise ratio (CNR). IR The IR technique (Fig. 1B) consisted of an inversion pulse, activated at the first heart beat after detection of the end-expiratory respiration phase, followed by an inversion time (TI) corresponding to 2 or 3 RRintervals in order to maximize contrast between viable and infarcted myocardium. A second trigger was initiated after the previously set TI and a segmented gradient echo sequence (turbo-FLASH) with Cartesian k-space sampling was acquired immediately after the R-wave. Imaging parameters included TR ¼ 5 msec, TReff ¼ respiration cycle (1500 msec), TI ¼ 2 or 3 RR-interval (300–400 msec), TE ¼ 1.5 msec, flip angle ¼ 20 , 1 average, 9 slices, 1 frame, and scan time of 15 6 5 minutes. The inversion pulse consisted of a nonselective adiabatic 180 RF pulse of 8 msec. Four phase encoding steps were used to reduce acquisition time while maintaining the acquisition window short to minimize motion effects.
Protti et al.
was followed by a segmented gradient echo acquisition (turbo-FLASH) with Cartesian k-space sampling. One k-space line per frame per heartbeat was acquired. Look–Locker parameters included TR ¼ RR-interval, TReff 2 seconds, TE ¼ 1.5 msec, flip angle ¼ 20 , 3 averages, 1 slice, 10 frames, and scan time 15 minutes. Look–Locker T1 values were used in this article after following correction over the application of RF pulses (17). Simulations SNR and CNR were estimated by solving the Bloch equations for IR-turbo-FLASH and cine-FLASH for infarcted and viable myocardium as well as for blood. Mz of the imaging sequence was calculated according to: MzðpostÞ ¼ M0 ½M0 MzðpreÞcosðaÞexpðTR=T 1Þ ½1 with TR being the repetition time, a the imaging flip angle, and T1 the relaxation time of either infarct, normal myocardium, or blood. To account for relaxation between the last a pulse of the imaging sequence and the IR prepulse, TR was modified to TR ¼ RR interval – TI – AQ for the last flip angle of the IRturbo-FLASH sequence. For cine FLASH, steady state was simulated by recursive programming of Mz. Imaging parameters were identical for simulations and in vivo experiments. Look–Locker-corrected T1 tissue values were used. For cine-FLASH, inflow of fresh blood was taken into account when simulating Mz of blood from Eq. [1]. Image Analysis SNR and CNR From Experimental Data SNR and relative CNR of CMR images was assessed using ImageJ (NIH, Bethesda, MD) for all experiments. Regions of interests (ROIs) were selected within Gdenhanced areas, viable myocardium, and blood. Mean signal intensity was estimated from such ROIs and noise was determined as standard deviation (SD) in the same ROI. SNR and relative CNR between tissues and blood (for SI>0) was expressed as a percentage and calculated as (9): SNR ¼ SItissue =Noise CNR ¼ ððSItissue1=Noise SItissue2=NoiseÞ =SItissue1=NoiseÞ 100%
½2 [3]
SNR and CNR for simulated data were calculated accordingly. Postmortem Analysis of Infarct Size
Look–Locker The optimal TI for nulling viable myocardium was determined using a Look–Locker sequence (16). The sequence consisted of an adiabatic inversion pulse performed immediately after the detection of the first heart beat after the respiratory trigger. The prepulse
All animals were culled by cervical dislocation under general anesthesia. Then the thorax was opened and the heart excised and briefly washed in ice-cold saline. The atria and right ventricle were separated from the LV, which was then placed in a Cryotube and rapidly cooled (to facilitate sectioning) by freezing
LGE of Acute MI
881
at 80 C for 20 minutes. Subsequently, the LV was cut manually using a scalpel into 4–5 transverse sections. These sections were rapidly transferred to a solution of 1.5% triphenyltetrazolium chloride (TTC) in a buffer solution of Na2HPO4/NaH2PO4 at pH 7.4 and temperature 37 C. The sections were kept in TTC solution at this temperature for 20 minutes. They were then transferred to a solution of 10% formalin (in phosphate-buffered saline) and left at room temperature overnight. Sections were imaged using a flatbed scanner (CanoScan LiDe 200, Canon), with the area of infarction as a proportion of the area of the whole section measured using Scion Image (Frederick, MD). This was performed for both sides of the section and the mean percent infarct for that section was calculated. This value was then multiplied by the (mass of that section / total mass of LV) to take into account the different masses of each section, producing the %LV infarct represented by that section. By summing this value derived from each section from one heart, the total %LV infarct for that heart was calculated. Infarct Area Analysis by CMRI Infarct size analysis was performed using a semiautomated in-house developed computer software program. The software was able to delineate the infarct region based on myocardial pixel intensities relaying on a threshold-based approach, thus minimizing user input to generate an ROI. The software generated a cluster of pixels encompassing LV infarcted tissues due to the different signal intensity with remote myocardium and blood. The same approach was applied to viable myocardium. Infarct, remote, and total LV areas were then automatically estimated and introduced in Eq. [4]. At 2 days post-MI, the myocardial wall did not demonstrate significant reduction in thickness compared to a control animals and no hypertrophy was seen either. Hence, the infarct area percentage was calculated as follows: Infarct area ¼
base X apex
areainf areainf
EF ¼ SV =EDV 100
½6
expressed in percent. CO ¼ SV heart rate
½7
expressed in mL/min. LV mass ¼ gmyocardium
base X
ðepicardialvolume
apex
endocardialvolumeÞ slicethickness
[8]
expressed in g (19). The specific gravity (g) of the myocardium is 1.055 g/cm2 (18). The same approach was maintained for infarcted hearts. The aforementioned cardiac preclinical computer software automatically calculated all the parameters after detecting epicardial and endocardial LV borders for end diastolic and end systolic frames. Basically, a cluster of pixels is generated in the LV blood pool to define the endocardial border. Whenever such border was not correctly placed, manual correction was applied. Statistical Section Student’s t-test was performed for infarct areas, LV mass, EF SV, and CO for comparison of continuous variables using GraphPad Prism (San Diego, CA) software. P-values of
