Improved Detection of Subendocardial Hyperenhancement in Myocardial Infarction Using Dark Blood Pool Delayed Enhancement MRI

Cardiopulmonary Imaging Original Research Farrelly et al. Dark Blood Delayed Enhancement MRI in Myocardial Infarction Cardiopulmonary Imaging Original Research Cormac Farrelly 1 Wolfgang Rehwald 2 Michael Salerno 3 Amir Davarpanah 1 Aoife N. Keeling 1 Jason T. Jacobson 4 James C. Carr 1 Farrelly C, Rehwald W, Salerno M, et al. Keywords: cardiac MRI, dark blood pool delayed enhancement, delayed enhanced imaging, myocardial infarction DOI:10.2214/AJR.10.4418 Received February 8, 2010; accepted after revision June 29, 2010. W. Rehwald is an employee of Siemens Healthcare. He had no control regarding the inclusion of data and information that might have represented a conflict of interest. W. Rehwald and M. Salerno have patented the dark blood pool delayed enhancement technique. They had no control regarding the inclusion of data and information that might have represented a conflict of interest. 1 Department of Radiology, Cardiovascular Imaging Section, Northwestern University, 737 N Michigan Ave., Ste. 1600, Chicago, IL 60611. Address correspondence to C. Farrelly (farrellycormac@gmail.com). 2 Cardiovascular MR Center, Duke University Medical Center, Durham, NC. 3 Department of Cardiovascular Medicine, University of Virginia Health System, Charlottesville, VA. 4 Bluhm Cardiovascular Institute, Northwestern University Feinberg School of Medicine, Chicago, IL. AJR 2011; 196:339 348 0361 803X/11/1962 339 American Roentgen Ray Society Improved Detection of Subendocardial Hyperenhancement in Myocardial Infarction Using Dark Blood Pool Delayed Enhancement MRI OBJECTIVE. Delayed enhancement MRI using fast segmented k-space inversion recovery (IR) gradient-echo imaging is a well established bright-blood technique for identifying myocardial infarction and is used as the reference standard sequence in this study. The purpose of this study was to validate a recently developed dark blood pool delayed enhancement technique in a porcine animal model, evaluate its performance in human patients, and quantify its performance compared with the reference standard in both. SUBJECTS AND METHODS. In an animal study, the reference standard and dark blood pool delayed enhancement were assessed in three pigs with induced myocardial infarction. In a human study, 26 patients, 31 81 years old (19 men and seven women), with a known history of myocardial infarction were imaged using the reference standard and dark blood pool delayed enhancement. Contrast-to-noise ratio (CNR), signal intensity ratio, signal-to-noise ratio (SNR), and qualitative scores of hyperenhancement were recorded. Measurements were compared using paired samples t test and Wilcoxon s signed rank test. RESULTS. In the animal study, the mean CNR of infarct to blood pool was 11 times higher for dark blood pool delayed enhancement than for the reference standard. The mean SNR was 4.4 times higher for the reference standard. In the human study, the mean CNR and signal intensity ratio of hyperenhancing myocardium to the blood pool were 1.9 (p = 0.04) and 5.5 (p < 0.01) times higher, respectively, for dark blood pool delayed enhancement compared with reference standard. The mean CNR and signal intensity ratio of hyperenhancing myocardium to normal myocardium and SNR were 2.8 (p < 0.01), 1.3 (p = 0.07), and 2.8 (p < 0.01) higher, respectively, for the reference standard. Qualitative analysis identified seven extra segments with grade 1 scars using dark blood pool delayed enhancement (p < 0.01). CONCLUSION. Dark blood pool delayed enhancement is complementary to the reference standard. It can detect more subendocardial foci of hyperenhancement, thus potentially identifying more infarcts and changing patient management. C ardiovascular disease is the leading cause of mortality in the United States, with myocardial infarction causing one in every five deaths [1]. Detection of myocardial infarction and quantification of the extent and severity of myocardial injury is important because these are strong predictors of patient outcome [2]. Improved therapy is leading to increased survival after myocardial infarction, but these patients are often left with residual myocardial damage. Interventions that protect viable myocardium and reduce injury improve left ventricular function and longterm survival [2, 3]. In recent years, cardiac MRI has evolved into routine clinical practice, and applications of cardiac MRI continue to expand. It has been known for over 20 years that regions of acute myocardial infarction enhance after administration of gadolinium-based contrast agents [4 8]. Since these initial investigations, utilization of ECG-gated MRI, improved hardware with stronger gradients, and improved field homogeneity in combination with new pulse sequences has led to superior visualization of infarcted tissue. Simonetti et al. [9] initially reported the use of delayed enhancement breath-hold inversion recovery (IR) segmented turbo FLASH sequence for T1-weighted contrastenhanced imaging of myocardial infarction (Fig. 1). This IR fast segmented k-space gradient-echo technique (Fig. 1) offers high spatial resolution and can identify both acute and chronic myocardial infarction [10, 11]. AJR:196, February 2011 339

Farrelly et al. +M0 M0 Nonselective IR TI Normal Myocardium Infarct Readout of One Segment Blood (With Contrast Agent) Normal Myocardium Fig. 1 Delayed enhancement inversion recovery (IR) turbo FLASH (reference standard) sequence diagram and corresponding relaxation curves of normal myocardium (black solid line), infarct (red solid line), and blood with contrast agent present (dashed blue line). Fast segmented k-space gradient-echo technique involves single IR preparatory pulse played between 250 and 380 milliseconds before data readout. Exact timing depends on amount of gadolinium-based contrast agent administered and time after its administration, typically 5 30 minutes. Inversion time (TI) is adjusted so that relaxation curve of normal myocardium crosses zero or is nulled during reading out center of k-space. Thus, normal myocardium appears dark. Infarct recovers faster, has higher signal, and consequently appears bright in image. Note that there is little image contrast between infarct and blood; in other words, gap between infarct and blood curves is small. +M 0 = magnetization immediately before first preparation IR pulse. It can distinguish between reversible (viable myocardium) and irreversible myocardial injury [3, 12] and allows high contrast between infarct and normal myocardium. It has become the most widely used MR sequence to depict late-enhancing areas of myocardium and is considered by many to be the clinical reference standard for viability imaging [13 15]; as such, we will refer to it as the reference standard from here on. However, despite these strengths, the reference standard is a bright-blood technique; hence, hyperenhanced infarcted myocardium has signal intensity (SI) similar to that of blood in the ventricular chamber. There is, therefore, poor contrast between blood pool and subendocardial areas of hyperenhancement, which makes small foci of subendocardial enhancement potentially difficult to detect. The identification of small subendocardial regions of hyperenhancement in patients with known left ventricular dysfunction is important because it may differentiate ischemic from nonischemic cardiomyopathy. The diagnosis of ischemic cardiomyopathy will lead to a change in patient management, by a combination of lifestyle modification, medical therapy, or revascularization. Large population studies indicate that survivors of myocardial infarction, whether the infarct was symptomatic or not, have three to 14 times the mortality rate of the general population [16]. Several MR techniques designed to improve differentiation between infarct and blood pool [17 19] have been proposed, including 3D delayed enhanced imaging [17] and the use of cine images as a reference to outline the subendocardial border, which can then be compared with delayed enhanced images [18]. Alternatively, T2-weighted images can be acquired within the same breath-hold as T1-weighted delayed enhanced images [19]. Standard double-inversion dark-blood imaging [20], incorporating a nonselective followed by a slice-selective inversion pulse, does not work in conjunction with delayed enhancement because the contrast agent unfavorably alters the required timing. With these issues in mind, a new technique aimed at simultaneously nulling the blood pool and myocardium while maintaining high signal within enhancing myocardium has been described recently [21 23]. This dark blood pool late enhancement technique combines a slice-selective inversion pulse, followed by a precisely timed nonselective inversion pulse to effectively decouple the T1 relaxation curves of the infarcted myocardium and the blood pool and to allow simultaneous nulling of both the normal myocardium and the blood pool (Fig. 2). The purpose of this study was to validate the dark blood pool delayed enhancement technique in a porcine animal model, to evaluate its performance in human patients with known myocardial infarction in a clinical setting, and to quantify its performance compared with reference standard imaging in both. Subjects and Methods Animal Study The study initially involved cardiac MR evaluation of myocardial infarction in three porcine models. All animal studies were approved by the Northwestern University Animal Care and Use Committee. Animal Preparation and Imaging Three pigs underwent experimental infarction. Our modification of the closed chest infarction procedure developed by Eldar et al. [24] has been described elsewhere [25]. Using a percutaneous femoral approach, an 8-French AL1 or AL2 guide catheter was positioned in the left anterior descending coronary artery (one case) and in the left circumflex artery (two cases), and a 2 2.5-mm angioplasty balloon was advanced just distal to the second diagonal and the first obtuse marginal branches, respectively. Thirty seconds after balloon inflation, 300 µl of agarose gel beads (diameter, 75 150 µm; Bio-Rad Laboratories) diluted in 1.5 ml of saline was injected to permanently occlude the artery distal to the site of balloon occlusion. Thirty seconds later, the balloon was deflated and the catheter was withdrawn. The evolving infarction was assessed using continuous ECG and hemodynamic monitoring. The animal was maintained under general anesthesia until the arterial sheath was removed 30 minutes after infarction. Animals were housed for 72 hours after infarction before undergoing the MRI study. All animals were sedated using ketamine (15 20 mg/g intramuscularly [IM]), atropine (0.05 mg/kg IM), and xylazine (1 2 mg/kg IM). They were then intubated, placed on a 1.5-T scanner (Magnetom Espree, Siemens Healthcare), and ventilated with gas anesthesia (2 3% isoflurane and 1 L of oxygen/min at a rate of 12 15 breaths/min) using a standard anesthesia unit (2B, Dräger Narkomed). Normal saline (0.9%) was administered IV at the rate of 2 3 ml/kg/hr throughout the scan. Anesthesia was assessed at regular intervals not exceeding 340 AJR:196, February 2011

Dark Blood Delayed Enhancement MRI in Myocardial Infarction +M0 M0 Selective IR TD TR Nonselective IR TI Blood Readout of One Segment Infarct 15 minutes by direct observation within the magnet. The animals were imaged in a supine position using a 14 28-cm flexible six-element cardiac array coil. Standard scout and balanced steady-state free precession cine images (true FISP, Siemens Healthcare) were acquired in multiplanar short- and long-axis views. Ten minutes after IV administration of 0.2 mmol/kg gadopentetate dimeglumine (Magnevist, Bayer Healthcare Pharmaceuticals) an inversion time (TI) scout was performed as described elsewhere [26]. The technician then selected the optimal TI from the TI scout (Fig. 1) before obtaining a stack of shortaxis reference standard images. These were acquired to encompass the entire heart from base to apex, with an 8-mm section distance and a slice thickness of 6 mm. A second TI scout was then performed to allow the extra time from contrast injection. The two optimal TIs required to null the myocardium and the blood pool were selected (Fig. 2). Dark blood pool delayed enhancement imaging was then performed from base to apex again with an 8-mm section distance and a Blood (With Contrast Agent) Normal Myocardium Fig. 2 Dark blood pool delayed enhancement. Graph depicts relaxation curves resulting from selective inversion recovery (IR) and later nonselective IR pulse for normal myocardium (black solid line), infarcted myocardium (red solid line), and blood (blue dashed line) 5 15 minutes after IV injection of contrast agent. Because of correctly chosen spacing, curves of blood and normal myocardium cross zero at same time (solid black arrow). Magnetization in infarct has recovered faster and appears bright in image. Rather than specifying one inversion time (TI) to null normal myocardium, as is done in classic delayed enhancement imaging, user provides two TIs to null both normal myocardium and blood. Spacing between inversion pulses (time delay [TD]) is then automatically calculated by sequence algorithm and protocol parameters TR and TI are adjusted accordingly. Time from nonselective IR pulse to center of k-space (TI blood ) is chosen to null blood and only depends on its T1. TD is set to null normal myocardium when blood is nulled (i.e., TD is chosen so that combination of both IR pulses will also null myocardium). Note that because infarcted myocardium experiences both IR pulses but blood pool only experiences second pulse, infarct will not be nulled even if T1 of infarct and blood pool are similar. +M 0 = equilibrium magnetization. Magnetization before first preparation IR pulse is close to but smaller than +M 0, its exact value depending on patient s R-R interval and trigger pulse used. slice thickness of 6 mm. Each image was acquired during a breath-hold of approximately 8 12 seconds and was ECG gated. On completion of imaging, animals were killed and the heart was removed. Ex vivo specimens of the pig hearts were sectioned into 10-mm slices. This was done in a short-axis plane to correspond with short-axis MR images. Gross anatomic examination was performed by a cardiac pathologist. In addition, histologic analysis was performed after staining with 2% triphenyltetrazolium chloride. Images were analyzed, and the regions of hyperenhancement on MRI were compared with the histologic specimens. One representative slice position with hyperenhancement present on reference standard was selected. This image and the corresponding dark blood pool delayed enhancement image at an identical slice position were subsequently used for quantitative measurements in data analysis. Patient Study Authors who are not employees of Siemens Healthcare controlled the inclusion of all data and information that might have represented a conflict of interest for one of the coauthors, who is an employee of that company. All patient studies were approved by the Northwestern University institutional review board and were HIPAA compliant. Written informed consent was waived by the review board because the dark blood pool delayed enhancement technique incorporates novel timing algorithms for well-established MRI pulse sequences and required no extra contrast agent to be administered. Patient Population Twenty-six consecutive patients (19 men and seven women; mean [± SD] age, 60 ± 13 years; range, 31 81 years) with a known history of myocardial infarction, referred to our institution for routine clinical MRI, were imaged between November 1, 2008, and March 31, 2009. None had unstable angina, New York Heart Association class 4 heart failure, or contraindications to MRI, such as a cardiac pacemaker or an automatic implantable cardiac defibrillator. All 26 studies were included for data analysis. Prior myocardial infarction was clinically confirmed as follows: 19 patients had a documented troponin leak in our institution. Mean peak serum levels were 19.8 ± 40.5 ng/ml (range, 0.05 157.7 ng/ml; normal range, 0 0.03 ng/ml). The time between troponin leak and MRI ranged from 24 hours to 49 months. All seven patients without a documented troponin leak in our institution had a prior admission to an outside institution where they were given a diagnosis of myocardial infarction. Three of these patients had Q waves on their ECG. The remaining four had significant coronary artery disease (at least a 70% stenosis) on digital subtraction coronary angiography performed at our institution. All 26 patients had subendocardial enhancement on reference standard images. Dark Blood Pool Delayed Enhancement Pulse Sequence Figure 2 shows the typical relaxation curves and illustrates the timing parameters of the dark blood pool delayed enhancement pulse sequence. As with the reference standard [9], the pulse sequence requires adaptation of a TI on a case-by-case basis to achieve optimal signal contrast between infarcted and normal myocardium. Fourteen to 20 minutes after administration of 0.2 mmol/kg of gadopentetate dimeglumine, a TI scout was performed in the standard manner [26]. This involved a segmented IR cine true FISP pulse sequence at a midventricular short-axis location and acquisition of images with TI values of 185 515 milliseconds. Parameters used were a TR/ TE of 2.2/1.1 milliseconds and a flip angle of 50. The spatial resolution was typically 2.6 1.8 10 mm, and the matrix was 192 115. AJR:196, February 2011 341

Farrelly et al. Unlike for the reference standard, the controller or technician now selects two TIs from the TI scout instead of just one. The first is the TI needed to null the blood pool (TI blood ), which is the TI on the scout at which the blood-pool signal is at its lowest (Fig. 2). The second is the TI needed to null the normal myocardium (TI normal ) that is, the TI on the scout at which normal myocardial signal is at its lowest as with the reference standard. These two TI values are fed into the user interface on the scanner. Note that TI normal is not illustrated in Figure 2 because the nulling of normal myocardium at the same time as when blood is nulled is achieved by the combination of both IR pulses. Using the two user-provided TIs, the dark blood pool delayed enhancement algorithm then automatically calculates and sets the optimal time from the slice-selective IR pulse to the end of the data readout (referred to as TR on the scanner user interface) and the time from the nonselective IR pulse to the acquisition of the k-space center-line (parameter TI). The time between the first and the second inversion pulse (i.e., time delay [TD]) is calculated by describing the relaxation curves of Figure 2 as equations and solving for TD so that the relaxation curve of normal myocardium is nulled at the same time when blood is nulled, as follows: TD = TI 2M normal 1n 0 1n(2) M M 2 0 where M 0 is the magnetization immediately before the first preparation inversion pulse and M 2 is the magnetization following the second inversion pulse: TI blood 1n(2) M 2 = M 0 e TI normal 1 These equations can be reformulated to allow a magnetization of normal myocardium just above the null point to create dark-gray myocardium to provide better endocardial border definition. If the resulting timing does not fit into one R-R interval, the scanner operator can manually add a trigger delay (from the R wave on the ECG to the slice-selective inversion pulse) to extend the imaging over two R-R intervals. Note that the trigger delay for imaging with a single R-R interval should not exceed 50 milliseconds; otherwise, the blood suppression is compromised. For imaging with two R-R intervals, the trigger delay is much larger to move the slice-selective IR pulse into diastole. Typical pulse sequence parameters were as follows: 7.1/3.3 milliseconds; flip angle, 25 ; bandwidth, 140 Hz per pixel; and matrix, 256/156. The mean TI blood selected by the controller or technician for the blood pool was 258.1 ± 28.1 milliseconds (range, 210 300 milliseconds). The mean TI normal selected by the controller or technician was 309.2 ± 27.0 milliseconds (range, 260 360 milliseconds). To avoid unnecessarily complicating the user interface at the scanner, the TD computed by the dark blood pool delayed enhancement sequence algorithm is not displayed. Each image was acquired in a single breath-hold (< 17 seconds). Imaging Protocol All images were acquired on a 1.5-T unit (Avanto, Siemens Healthcare) using an eight-channel phased-array body coil. A 20-gauge cannula was placed in an antecubital vein. Patients were scanned in the supine position. Scout and cine images were acquired in standard long- and short-axis views; 0.2 mmol/kg of gadopentetate dimeglumine (Magnevist, Bayer Healthcare Pharmaceuticals) was then administered IV. At least 10 minutes were allowed before the reference standard sequence was used to cover the entire left ventricle in the short-axis view, with a 1-cm section distance and a slice thickness of 6 mm. Typical pulse sequence parameters were as follows: 7.1/3.3 milliseconds; flip angle, 25 ; bandwidth, 140 Hz per pixel; and matrix, 256/156. A dark blood pool delayed enhancement shortaxis image was then acquired at a slice position identical to that of one of the reference standard images encompassing a region of myocardial hyperenhancement. The representative section location was chosen at the scanner by two investigators (each with at least 6 years of radiologic experience) in agreement. The TI scout was repeated before acquiring the dark blood pool delayed enhancement image. This allowed adjustment of the optimal TI for myocardial nulling due to prolongation of tissue T1 with increasing delay from contrast injection. It also allowed selection of the optimal TI for nulling of the blood pool. The additional time needed to perform the second TI scout and acquire the dark blood pool delayed enhancement image was approximately 4 minutes. Data Analysis Quantitative analysis Quantitative analysis was performed for the animal and patient studies by two independent blinded observers (both with at least 6 years of radiologic experience). All images were randomized and presented for evaluation at a 3D workstation (Leonardo, Siemens Healthcare) and were analyzed by placing regions of interest (ROIs) in the infarct (hyperenhancement), the remote myocardium (no hyperenhancement), the left ventricular cavity blood pool (fixed ROI, 2 cm 2 ), and the background ventral and lateral to the patient or animal (fixed ROI, 2 cm 2 ). Noise was recorded as 1.5 times the SD of the SI within the background ROI [27, 28]. The following parameters were calculated for all reference standard and dark blood pool delayed enhancement images: First, the contrast-to-noise ratio (CNR) of myocardial hyperenhancement to the blood pool and to normal myocardium was calculated for all animal and patient studies [9], where CNR = (mean SI of the hyperenhanced region mean SI of the normal myocardium or mean of the SI blood pool) / 1.5 SD of background noise. Second, the SI ratio of the hyperenhanced myocardium to the blood pool and SI ratio of the hyperenhanced myocardium to normal myocardium were calculated for all patient studies as the mean SI of the hyperenhanced region divided by the mean SI of the blood pool or normal myocardium. Finally the signal-to-noise ratio (SNR) of the hyperenhancing myocardium was measured for all studies, calculated as the mean SI of the hyperenhanced region divided by 1.5 SD of background noise. These measurements were repeated 1 month later from the same images by one of the readers. Left ventricular ejection fractions were calculated from the short-axis cine sequences using semiautomated software (ARGUS, Siemens Healthcare). Because the dark blood pool delayed enhancement technique relies on effective blood exchange out of the imaging slice during acquisition, the ratio of signal in the blood pool compared with signal of normal (nonenhancing) myocardium was also compared between patients with left ventricular ejection fractions either greater than 40% or less than or equal to 40% for dark blood pool delayed enhancement as follows: SI of the blood pool was divided by the SI of the normal myocardium. Interobserver and intraobserver variability were calculated for all contrast and SNR measurements. The mean value between the two readers was used as the final value for the purpose of calculations and further statistical analysis. Qualitative analysis Qualitative analysis for all studies was performed by two independent blinded reviewers (both with at least 6 years of radiologic experience). For the animal study, a stack of reference standard and dark blood pool delayed enhancement short-axis images for each animal was presented for evaluation at a 3D workstation (Leonardo, Siemens Healthcare). The number of segments with hyperenhancement were recorded and graded according to the 17-segment model [29] for both techniques. For the patient study, a single representative short-axis view, determined by the representative section location chosen at the time of scanning, for both delayed enhancement techniques on each patient was randomized and presented for evaluation at the same 3D workstation. The single views for each enhancement technique were presented at separate times to avoid any bias that could be caused by side-by-side analysis or a memory 342 AJR:196, February 2011

Dark Blood Delayed Enhancement MRI in Myocardial Infarction effect. Hyperenhancement was again graded according to the 17-segment model with either six (base and midchamber) or four (apex) segments presented for evaluation on each patient. The transmurality of hyperenhancement in each segment was graded as follows: 0%, grade 0; 1 25%, grade 1; 26 50%, grade 2; 51 75%, grade 3; and 76 100%, grade 4. The number of segments with each grade of hyperenhancement was recorded. Readers also scored their confidence in grading the transmural extent of hyperenhancement for each image: 0, no confidence, cannot tell; 1, very little confidence; 2, significant doubt; 3, slight doubt; and 4, very confident. Wall motion was evaluated from short-axis cine sequences taken at slice positions identical to those of the delayed enhancement images on each patient. This was also graded regionally according to the 17-segment model. The severity of wall motion abnormality in each segment was graded according to a 5-point scale as follows: 0, no wall motion abnormality; 1, mild-to-moderate hypokinesia; 2, severe hypokinesia; 3, akinesia; and 4, dyskinesia. Finally, the short-axis reference standard and dark blood pool delayed enhancement images at identical slice positions for each patient were presented side by side (readers were therefore not blinded), and readers were asked to score conspicuity (i.e., determine on which image the presence of scar is more obvious). The readers were given the choices of reference standard, dark blood pool delayed enhancement, or hyperenhancement equally conspicuous on both. Total scar burden for the delayed enhancement sequences was calculated in each patient by adding the grade of transmural enhancement for each segment together. The total score for wall motion abnormality in each patient was reached by adding together the severity scores for each segment. These measurements were used for calculation of interobserver variability. All final scores for analysis were agreed by consensus. Statistical Analysis Statistical analysis was performed using commercially available software (SPSS, version 17.0; SPSS) for all patient studies. CNR and SNR measurements were compared between reference standard and dark blood pool delayed enhancement using paired samples t test. An independent samples t test was used to compare blood-pool SI to normal myocardium SI for patients with a left ventricular ejection fraction greater than 40% or less than or equal to 40%. Qualitative score data assessing the number and transmurality of segmental myocardial hyperenhancement between both delayed enhancement protocols were compared according to nonparametric TABLE 1: Quantitative Results for Patient Study Population: Dark Blood Pool Delayed Enhancement Sequence Compared With Reference Standard Ratio test of two related samples using Wilcoxon s signed rank test. The Mann-Whitney U test was used to compare total scar burden between two groups (wall motion abnormality scores > and 4) based on nonparametric test of two independent samples. In all statistical tests, a p value of less than 0.05 was considered significant. Interobserver and intraobserver variability for quantitative measurements of CNR, SI ratio, and SNR and interobserver variability for measurements of total wall motion abnormality, total scar burden, and confidence grading transmurality of hyperenhancement was determined by reliability statistics using intraclass correlation coefficient (ICC). Agreement of more than 0.41, 0.61, and 0.81 was considered moderate, good, and very good, respectively. Reference Standard Dark Blood Pool Delayed Enhancement a Contrast-to-noise ratio (of infarct to blood pool) 10.4 19.8 (190) 0.04 Signal intensity ratio (of infarct to blood pool) 1.9 10.5 (553) < 0.01 Contrast-to-noise ratio (of infarct to normal myocardium) Signal intensity ratio (of infarct to normal myocardium) 49.6 17.7 (36) < 0.01 13.3 10.2 (77) 0.07 Signal-to-noise ratio (of infarct) 55 19.6 (36) < 0.01 Note Reference standard is delayed enhancement inversion recovery turbo FLASH. a Numbers in parentheses are percentage of reference standard. Results Animal Study Hyperenhanced subendocardium and wall motion abnormality were present in the distribution of the infarcted territory on all three pig hearts for delayed enhancement reference standard and dark blood pool delayed enhancement images. The mean CNR of hyperenhanced myocardium to blood pool was 11 times higher for dark blood pool delayed enhancement than for the reference standard. The mean CNR of hyperenhanced myocardium to normal (nulled) myocardium was 4.3 times higher for the reference standard than for dark blood pool delayed enhancement. The SNR was 4.4 times higher for the reference standard. Seven, six, and six segments with grade 4 transmural hyperenhancement were recorded for each of the three porcine models using both reference standard and dark blood pool delayed enhancement. No segments had grade 1, 2, or 3 scars. No myocardial segment had hyperenhancement on one delayed enhancement technique that did not have it on the other, and neither technique showed enhancement where there was no histologic infarct. Human Study Quantitative results The mean CNR of hyperenhanced myocardium to the blood pool was 1.9 times higher for the dark blood pool delayed enhancement technique compared with the reference standard (p = 0.04). The mean SI ratio of the hyperenhanced myocardium to the blood pool was 5.5 times higher for dark blood pool delayed enhancement (p < 0.01). The mean CNR of infarcted area to the normal myocardium was 2.8 times higher for the reference standard than for dark blood pool delayed enhancement (p < 0.01). The mean SI ratio of the infarcted myocardium to normal myocardium was 1.3 times higher for the reference standard. This approached but did not reach statistical significance (p = 0.07). The SNR for the reference standard was 2.8 times higher than for dark blood pool delayed enhancement (p < 0.01). These results are summarized in Table 1. Using dark blood pool delayed enhancement, the mean SI ratio of blood pool to normal myocardium was 2.7 times higher in patients with a left ventricular ejection fraction less than or equal to 40% (n = 12) than in patients with an ejection fraction of greater than 40% (n = 14) (p = 0.04). Interobserver agreement was very good for reference standard CNR measurements compared with the blood pool and the normal myocardium (ICC, 0.82 and 0.92, respectively), dark blood pool delayed enhancement CNR measurements (ICC, 0.89 and 0.93, respectively), reference standard SI ratio measurements (ICC, 0.84 and 0.91, respectively), and dark blood pool delayed enhancement SI p AJR:196, February 2011 343

Farrelly et al. Fig. 3 Illustration of number of segments in which grade 1 4 hyperenhancement was detected. One hundred forty segments were analyzed in 26 patients. Using dark blood pool delayed enhancement (gray bars), seven more segments with grade 1 scar and one extra segment with grade 4 scar were detected compared with reference standard (i.e., delayed enhancement inversion recovery turbo FLASH; black bars). No. of Segments ratio measurements (ICC, 0.89 and 0.95, respectively). SNR measurements for reference standard and dark blood pool delayed enhancement images also had very good agreement (ICC, 0.94 and 0.91, respectively). Intraobserver variability was very good for all CNR, SI ratio, and SNR measurements (ICC > 0.81). Qualitative results In total, 140 myocardial segments were evaluated. Of these, hyperenhancement was detected in 60 segments (42.9%) using reference standard and in 68 segments (48.6%) using dark blood-pool delayed enhancement (Fig. 3). All segments containing hyperenhancement on reference standard images also had hyperenhancement on dark blood pool delayed enhancement images. Qualitative assessment of hyperenhancement detected 10 segments (7.1%) with grade 1 scar (1 25% subendocardial hyperenhancement) using reference standard versus 17 segments (12.1%) with dark blood pool delayed enhancement. This difference was statistically significant (p = 0.04). The additional extra segment of hyperenhancement detected with dark blood pool delayed enhancement was a segment of grade 4 scar. Qualitative assessment of hyperenhancement detected 27 segments (19.3%) with grade 4 scar (75 100% subendocardial hyperenhancement) using reference standard versus 28 segments (20%) with dark blood pool delayed enhancement. There was no statistically significant difference in the two delayed enhancement techniques for detecting grade 2, 3, or 4 scars (p > 0.05 for each). A no-reflow zone, indicating acute microvascular obstruction, was not present on any of the patient images. The seven extra segments of grade 1 scar detected with dark blood pool delayed enhancement were in four patients. One of these patients, a 51-year-old man, had a coronary angiogram and a coronary stent placed in his proximal left circumflex artery 1 day after his cardiac MRI. This patient also had 1 segment of grade 2 and 1 segment of grade 3 scar on his cardiac MRI. The interventional cardiologist who performed the stenting reported that the segment of grade 1 scar detected with dark blood pool delayed enhancement helped him when deciding whether to perform subsequent intervention. It led him to conclude that the patient was more likely to have viable myocardium and could therefore benefit from stenting. This patient had a left ventricular ejection fraction of 32% calculated on his MRI. An echocardiogram performed 1 day before his MRI revealed an ejection fraction of 31%, and an echocardiogram 6 months later revealed an ejection fraction of 40% with mild improvement in lateral wall contraction. This patient did not undergo a follow-up MRI. The other three patients with extra grade 1 scar detected with dark blood pool delayed enhancement did not undergo cardiac intervention at 9, 9, and 5 months follow-up, respectively. One patient had enhancement in an anterolateral papillary muscle detected with dark blood pool delayed enhancement but not with reference standard. This extra focus of papillary muscle hyperenhancement was not included in the study analysis because it did not conform to a specific myocardial segment. This patient had mild mitral regurgitation but no flail leaflet on cardiac MRI or echocardiography. He is being monitored with serial echocardiography. An echocardiogram at 6 months revealed no change in mitral valve function. Conspicuity scores showed improved conspicuity for dark blood pool delayed enhancement images in 11 patients 14 cases where the hyperenhancement was equally obvious on dark blood pool delayed enhancement and reference standard and one where it was more conspicuous on reference standard. Readers were more confident grading transmurality of hyperenhancement on reference standard than on dark blood pool delayed enhancement images, with mean scores of 3.7 and 3.4, respectively. This difference was statistically significant (p = 0.04). Qualitative analysis of the cine sequences of all 140 segments showed wall motion abnormality in 66 segments. Of these 66 segments, 54 (81.8%) had hyperenhancement detected with reference standard. Six segments had hyperenhancement on reference 30 25 20 15 10 5 0 Grade 1 Grade 2 Grade 3 Grade 4 standard and no wall motion abnormality. Of the 66 segments with wall motion abnormality, 59 (89.4%) had hyperenhancement on dark blood pool delayed enhancement images. Nine segments had hyperenhancement on dark blood pool delayed enhancement and no wall motion abnormality. Interobserver agreement was very good and good for measurements of total scar burden and for confidence grading transmurality of hyperenhancement from reference standard images (ICC, 0.9 and 0.74, respectively). Interobserver agreement was very good and good for measurements of total scar burden and for confidence grading transmurality of hyperenhancement from dark blood pool delayed enhancement images (ICC, 0.86 and 0.71, respectively). Agreement grading wall motion abnormality was good (ICC, 0.81). Discussion The purpose of the animal portion of the study was to determine whether the dark blood pool delayed enhancement technique revealed enhancement in infarcted myocardium and to establish whether enhancement or artifact was incorrectly perceived as infarct in normal myocardium. Dark blood pool delayed enhancement sequences resulted in successful nulling of the blood-pool and correct identification of myocardial infarction, as determined by agreement with the reference standard technique and pathologic correlation. There was resultant superior contrast of hyperenhanced myocardium relative to the blood pool, with relatively well-preserved SNR and CNR between nonenhancing and hyperenhancing myocardium compared with the reference standard images (Fig. 4). On the basis of these promising results, dark blood pool delayed enhancement 344 AJR:196, February 2011

Dark Blood Delayed Enhancement MRI in Myocardial Infarction A Fig. 4 Short-axis images of porcine heart, obtained with the reference standard (i.e., delayed enhancement inversion recovery turbo FLASH; A) or dark blood pool delayed enhancement (B), 72 hours after induction of myocardial infarction in distribution of left anterior descending artery. Hyperenhancement of anteroseptum and anterior walls are present. Note that no-reflow zone, indicating microvascular obstruction, involving anterior wall (curved arrows) could be mistaken for thrombus on reference standard image. No-reflow zone involving anteroseptum (straight arrows) could be mistaken for grade 1 scar (involving < 25% of wall thickness) on reference standard image (A). On dark blood pool delayed enhancement image (B), boundaries between hyperenhanced myocardium and blood pool, for both right and left ventricle, are clearly defined, enabling definite diagnosis of no-reflow to be made. A Fig. 5 Images of 72-year-old man, obtained with the reference standard (i.e., delayed enhancement inversion recovery turbo FLASH; A) or dark blood pool delayed enhancement (B), 4 months after documented troponin rise. There is grade 1 subendocardial hyperenhancement (curved arrow; B) showing involvement of inferolateral wall on dark blood pool delayed enhancement image that is not evident on corresponding reference standard image (A). was performed on 26 patients with a diagnosis of prior myocardial infarction undergoing routine clinical cardiac MRI. The mean CNR of left ventricular myocardial hyperenhancement compared with the blood pool was 1.9 times higher for the dark blood pool delayed enhancement images than for reference standard, whereas the SI ratio of infarcted myocardium compared with the blood pool was more striking, being 5.5 times higher for dark blood pool delayed enhancement. Despite this favorable contrast performance of dark blood pool delayed enhancement, its SNR was 2.8 times lower than that of the reference standard in this study. The SNR of dark blood pool delayed enhancement is lower because the imaged slice experiences two IR pulses per imaging period and thereby suppresses a larger range of T1 species compared with the reference standard, which uses only a single IR pulse. Another factor leading to the lower SNR with the dark blood pool delayed enhancement could be a lower concentration of contrast media resulting from the longer delay, approximately 4 minutes, from the time of injection. The reason for the higher SI ratio of the hyperenhanced myocardium relative to normal nulled myocardium with reference standard is likely because, to null the myocardium with dark blood pool delayed enhancement, the time between the slice-selective inversion B B pulse and the nonselective inversion pulse must be precise to allow the combination of both radiofrequency pulses to null the myocardium. The large difference in CNR between hyperenhanced myocardium and the blood pool for dark blood pool delayed enhancement and reference standard that we observed in pigs versus humans (CNR 11 times higher for dark blood pool delayed enhancement than reference standard in pigs vs 1.9 times higher in humans) may be partially because imaging animal models in wellcontrolled circumstances often affects the quality of images obtained. Geometric differences such as the relatively small size of a pig s heart compared with a human s may also have made a difference. More foci of grade 1 enhancement (hyperenhancement involving 1 25% of mural thickness) were identified with the dark blood pool delayed enhancement technique (Figs. 3 and 5). This is likely because, on standard delayed enhancement techniques such as the reference standard, the blood pool has almost the same SI as enhancing subendocardium. Therefore, if the hyperenhancement does not involve a large volume of myocardium, it may be indistinguishable from the blood pool. The low SI of the blood pool on dark blood pool delayed enhancement obviates this problem and improves visualization of small subendocardial foci of enhancement. Numerous studies have shown that the transmural extent of hyperenhancement, using conventional delayed enhancement techniques such as reference standard, before revascularization is significantly related to the likelihood of improvement in segmental function and global ejection fraction after revascularization [3, 12, 14, 30]. Thus, identification of small regions of subendocardial hyperenhancement using dark blood pool delayed enhancement may not only aid in diagnosis but also potentially identify segments that would benefit from improved function after revascularization. This study was not specifically designed to evaluate whether the use of dark blood pool delayed enhancement could alter patient outcome but rather to investigate the feasibility of this new MRI technique in revealing myocardial infarction and to compare it to the more established reference standard technique. However, it is interesting that clinical management was altered in two of the 26 patients because of findings with dark blood pool delayed enhancement that were not evident on the reference standard. One had follow-up serial echocardiography to monitor AJR:196, February 2011 345

Farrelly et al. A Fig. 6 Short-axis images of 43-year-old man, obtained with the reference standard (i.e., delayed enhancement inversion recovery turbo FLASH; A) or dark blood pool delayed enhancement (B), 36 hours after documented troponin rise. Quantification of transmural extent of hyperenhancement involving anteroseptum is easier with reference standard because nonenhancing septal myocardium (curved arrow, A) stands out against high signal in right ventricular blood-pool compared with dark blood pool delayed enhancement image (B) where both blood-pool and normal myocardium are nulled. Note hyperenhancement involving anterolateral papillary muscle on dark blood pool delayed enhancement image (arrow, B) that is not as evident on reference standard image (A). Patient did not have flail mitral valve leaflet. Images illustrate complementary nature of reference standard and dark blood pool delayed enhancement imaging sequences. a focus of papillary muscle hyperenhancement identified with dark blood pool delayed enhancement. Another went on to have coronary stenting of his left circumflex artery with improved regional wall motion on follow-up echocardiography. This finding highlights the importance of detecting grade 1 scar because it suggests that the surrounding myocardium deep to the enhancement represents viable myocardium that may benefit from revascularization. It is reasonable to postulate that, if dark blood pool delayed enhancement can identify additional segments of grade 1 hyperenhancement in patients with known infarcts, it may also have the potential to identify small subendocardial infarcts in patients in whom conventional delayed enhancement techniques do not. This warrants investigation in further studies. Measurement of the spatial extent of ventricular wall motion abnormality alone can lead to an overestimation [11, 31] or an underestimation [11, 32] of infarct size. Using dark blood pool delayed enhancement, 13% of regions with hyperenhancement had no corresponding wall motion abnormality, and 10.6% of regions with abnormal wall motion had no detectable hyperenhancement. The potential of dark blood pool delayed enhancement images to identify subendocardial regions of hyperenhancement where there is no regional wall motion abnormality is important because subendocardial infarcts represent areas at risk for future infarction in the infarct-related arterial territory [33]. The SI in the blood pool, compared with that in the nulled nonenhancing myocardium, was 2.7 times lower when the left ventricular ejection fraction was greater than 40% compared with when it was less than or equal to 40% (p = 0.04). This reflects superior nulling of the blood pool when there is increased blood replacement between the application of the slice-selective and the nonselective inversion pulses. We propose that the ability to detect small subendocardial (grade 1) regions of hyperenhancement is more important clinically when there is little wall motion abnormality globally because it is in these patients that the diagnosis of small myocardial infarcts is particularly difficult. Patients with pronounced wall motion abnormality due to prior myocardial infarction are more likely to have large regions of readily identifiable myocardial hyperenhancement. Conspicuity scores were higher for dark blood pool delayed enhancement than for reference standard, indicating that infarct hyperenhancement subjectively stood out more on the dark blood pool delayed enhancement image. A region of grade 4 hyperenhancement involving the inferoseptum at the midchamber level identified with both techniques was read as extending into anteroseptum with the dark blood pool delayed enhancement technique. This represented the extra segment of grade 4 hyperenhancement recorded with dark blood pool delayed enhancement. The increased confidence that readers reported when grading the transmural extent of myocardial hyperenhancement with reference standard is understandable when one considers enhancement of the interventricular septum, for example. The dark blood pool delayed enhancement algorithm allows the magnetization of normal myocardium to be just above the null point. This creates darkgray myocardium while completely nulling the blood pool in an attempt to provide better endocardial border definition. Despite this, with dark blood pool delayed enhancement, the normal nonenhancing septum and the blood pool in both ventricles will all be low signal, making the amount of septal enhancement relatively difficult to quantify compared with reference standard (Fig. 6). As with conventional delayed enhancement techniques such as the reference standard, the TI of normal myocardium will change with time [34]. This effect is known to be most pronounced in the first 5 minutes after administration of contrast material and to trend toward a plateau with time. Similarly with dark blood pool delayed enhancement, the TI of blood would also be expected to change rapidly within the first 5 minutes and trend toward a plateau after 10 minutes. Although it is recognized that changes in TI with time may affect nulling of the blood pool, and it may have been helpful to document how fast the optimal TI changes by measuring the TI at different time points from multiple TI scout scans over a period of time, these changes would be expected to be minimal at the time point when dark blood pool delayed enhancement scanning occurred (i.e., 15 20 minutes after contrast administration). Anecdotally, operators noted little drift in TI blood when scanning. The use of phased-array surface coils in this study may have had some effect on the accuracy noise measurements, although the same measurement technique was used for each sequence, thus minimizing significant differences. SD of infarcted and normal myocardium was not recorded, although this may have been useful. This study has other limitations. Instead of the entire infarction, one representative short-axis view in each patient was selected for evaluation of the two different pulse sequences. This was done in an attempt to obtain reference standard and dark blood pool delayed enhancement images close together in time after contrast injection. This allowed more accurate comparison of quantitative measurements between both techniques. If a full stack of short-axis images had been B 346 AJR:196, February 2011