Abstract 4369860: Deep Learning Segmentation for Automated Measurement of Infarct Size in Preclinical Myocardial Infarction Models

Infarct size (IS) is the most robust end point for evaluating the success of preclinical studies on cardioprotection. The gold standard for IS quantification in ischemia/reperfusion (I/R) experiments is triphenyl tetrazolium chloride (TTC) staining, typically done manually. This study aimed to determine if automation through deep learning segmentation is a time-saving and valid alternative to standard IS quantification. High-resolution images from TTC-stained, macroscopic heart slices were retrospectively collected from pig experiments (n = 390) with I/R without/with cardioprotection to cover a wide IS range. Existing IS data from pig experiments, quantified using a standard method of manual and subsequent digital labeling of film-scan annotations, were used as reference. To automate the evaluation process with the aim to be more objective and save time, a deep learning pipeline was implemented; the collected images (n = 3869) were pre-processed by cropping and labeled (image annotations). To ensure their usability as training data for a deep learning segmentation model, IS was quantified from image annotations and compared to IS quantified using the existing film-scan annotations. A supervised deep learning segmentation model based on dynamic U-Net architecture was developed and trained. The evaluation of the trained model was performed by fivefold cross-validation (n = 220 experiments) and testing on an independent test set (n = 170 experiments). Performance metrics (Dice similarity coefficient [DSC], pixel accuracy [ACC], average precision [mAP]) were calculated. IS was then quantified from predictions and compared to IS quantified from image annotations (linear regression, Pearson's r; analysis of covariance; Bland-Altman plots). Performance metrics near 1 indicated a strong model performance on cross-validated data (DSC: 0.90, ACC: 0.98, mAP: 0.90) and on the test set data (DSC: 0.89, ACC: 0.98, mAP: 0.93). IS quantified from predictions correlated well with IS quantified from image annotations in all data sets (cross-validation: r = 0.98; test data set: r = 0.95) and analysis of covariance identified no significant differences. The model reduced the IS quantification time per experiment from approximately 90min to 20s. The model was further tested on a preliminary test set from experiments in isolated, saline-perfused rat hearts with regional I/Rwithout/with cardioprotection (n = 27). There was also no significant difference in IS between image annotations and predictions, but the performance on the test set data from rat hearts was lower (DSC: 0.66, ACC: 0.91, mAP: 0.65). IS quantification using a deep learning segmentation model is a valid and time-efficient alternative to manual and subsequent digital labeling.

Effect of Anemia and Red Blood Cell Transfusion in Acute Myocardial Infarction.

REPLY: Dr. Lygate (5) has suggested that the variability observed in infarct size in the surgical mouse myocardial infarction model (10 –70%) (1, 7) mandates larger sample sizes than used in our study (9). Although we agree that variable infarct size certainly occurs with this model, our presented data show clear, statistically significant differences with the given sample sizes. Our protocol may reduce variability in infarct size by careful attention to a proximal ligation of the left coronary immediately at the site where it emerges from under the left atrium as well as by eliminating mice that do not have a clear area of distal blanching of myocardium after immediate ligation (4, 8, 9). The more distal the ligation is made, the more the variability in coronary anatomy of the mouse heart impacts infarct size. Thus the referenced study by Patten et al. (7), which stated that the suture on the left coronary was placed approximately midway between the apex and base of the left ventricle, likely introduced greater variability in infarct size than we observed. Another study referenced by Dr. Lygate further makes this point in that Takagawa et al. (10) systematically created large infarctions by using proximal left coronary ligations (4 mm from apex) and small infarctions by distal ligations (2 mm from apex). We conclude that with careful attention to surgical technique, the range of infarct sizes can be much smaller than the 10 –70% suggested by Dr. Lygate and that with the techniques used in our study, statistically significant results with the stated sample sizes were obtained. Second, Dr. Lygate questions the validity of the infarct size measurements made at the mid-papillary muscle level. This approach is a common technique used in the literature, including a contemporary study (2). Although we certainly concede that absolute quantification of size of the infarction relative to the total left ventricle would require additional sampling, our purpose was to compare two groups of animals assessed in the same fashion at the mid-papillary level. This was the region where the samples were taken for assessment of apoptosis, so it was desirable to have the closest matched regions for apoptosis and infarct size measurements. Using this uniform sampling method, we observed statistically significant differences. Finally, Dr. Lygate raises the question as to whether the decrease in apoptosis, hypertrophy, and fibrosis were direct effects of ESC treatment or simply reflected the reduction in infarct size. Dr. Lygate states that the extent of apoptosis correlates directly with infarct size and cites relevant literature (3, 7); however, these references simply look at apoptosis levels following infarction without intervention. A broader examination of the literature in which intervention groups are compared with control myocardial infarction groups reveals that a correlation between infarct size and the extent of apoptosis is oftentimes not present (2, 6). To begin to address this issue directly for apoptosis in our study, we provided data in Figure 1H (9) showing that there is not a simple correlation between infarct size and apoptotic nuclei and that the two groups of data are clearly separate. Nevertheless, we acknowledge that fully understanding the mechanisms underlying the observed effects will require substantial future studies.

Purpose: To validate a computer algorithm for measuring myocardial infarct size on gadolinium enhanced MR images. The results of computer infarct sizing are studied on phase-sensitive and magnitude imaging against a histopathology reference. Materials and Methods: Validations were performed in 9 canine myocardial infarctions determined by triphenyltetrazolium chloride (TTC). The algorithm analyzed the pixel intensity distribution within manually traced myocardial regions. Pixels darker than an automatically determined threshold were first excluded from further analysis. Selected image features were used to remove false positive regions. A threshold 50% between bright and dark regions was then used to minimize partial volume errors. Post-processing steps were applied to identify microvascular obstruction. Both phase sensitive and magnitude reconstructed MR images were measured by the computer algorithm in units of % of the left ventricle (LV) infarction and compared to TTC. Results: Correlations of MR and TTC infarct size were 0.96 for both phase sensitive and magnitude imaging. Bland Altman analysis showed no consistent bias as a function of infarct size. The average error of computer infarct sizing was less than 2% of the LV for both reconstructions. Fixed intensity thresholding was less accurate compared to the computer algorithm. Conclusions: MR can accurately depict myocardial infarction. The proposed computer algorithm accurately measures infarct size on contrast-enhanced MR images against the histopathology reference. It is effective for both phase-sensitive and magnitude imaging.

Reply to “Letter to the editor: Infarct size measurements are critically important when comparing interventions affecting ventricular remodeling”

<i>Circulation: Cardiovascular Interventions</i> Editors’ Picks

BackgroundThere is a growing interest in developing non-invasive imaging techniques permitting infarct size (IS) measurements in mice. The aim of this study was to validate the high-resolution rodent Linoview single photon emission computed tomography (SPECT) system for non-invasive measurements of IS in mice by using a novel algorithm independent of a normal database, in comparison with histology.MethodsEleven mice underwent a left coronary artery ligature. Seven days later, animals were imaged on the SPECT 2h30 after injection of 173 ± 27 MBq of Tc-99m-sestamibi. Mice were subsequently killed, and their hearts were excised for IS determination with triphenyltetrazolium chloride (TTC) staining. SPECT images were reconstructed using the expectation maximization maximum likelihood algorithm, and the IS was calculated using a novel algorithm applied on the 20-segment polar map provided by the commercially available QPS software (Cedars-Sinai Medical Center, CA, USA). This original method is attractive by the fact that it does not require the implementation of a normal perfusion database.ResultsReconstructed images allowed a clear delineation of the left ventricles borders in all mice. No significant difference was found between mean IS determined by SPECT and by TTC staining [37.9 ± 17.5% vs 35.6 ± 17.2%, respectively (P = 0.10)]. Linear regression analysis showed an excellent correlation between IS measured on the SPECT images and IS obtained with TTC staining (y = 0.95x + 0.03 (r = 0.97; P < 0.0001)), without bias, as demonstrated by the Bland-Altman plot.ConclusionOur results demonstrate the accuracy of the method for the measurement of myocardial IS in mice with the Linoview SPECT system.

This study aims to explore the role and the mechanism of Parkin protein in cardiac function and ventricular remodeling in myocardial infarction (MI) rats, and to provide a new sight for the treatment of myocardial infarction. Fifty Sprague- Dawley (SD) male rats were randomly divided into 5 groups: sham operation group (Sham group), model group (MI group), low-dose Parkin group (L-Parkin group), middle-dose Parkin group (M-Parkin group) and high-dose Parkin group (H-Parkin group). The rat model of myocardial infarction was established by ligation of the anterior descending branch. Small animal ultrasound was used to measure cardiac function. The myocardial infarct size was observed by triphenyltetrazolium chloride (TTC) staining. The pathological changes of myocardial tissues were observed by hematoxylin-eosin (HE) staining. The myocardial cell apoptosis was detected by TUNEL assay. The mRNA expression of matrix metalloproteinase 2 (MMP2), matrix metalloproteinase 9 (MMP9), tissue inhibitor of matrix metalloproteinase 1 (TIMP1), tissue inhibitor of matrix metalloproteinase 2 (TIMP2) were detected by qRT-PCR. The expression of Parkin protein in myocardial tissue of rats was detected by Western-blot. Compared with MI group, left ventricular end-systolic volume (LVESV) and left ventricular end-diastolic volume (LVEDV) in Parkin overexpressing group were significantly decreased (p<0.05), while the value of left ventricular short axis shortening (FS) and left ventricular ejection fraction (EF %) in Parkin overexpression group were significantly increased (p<0.05). Overexpression of Parkin improved abnormal structure of myocardial tissue, reduced the size of myocardial infarct, made the arrangement of myocardium fibers more neatly and made the stain of myocardial cells more uniformly. Apoptosis index (AI) values were significantly decreased (p<0.05), and MMP2, MMP9, TIMP1 and TIMP2 mRNA levels were significantly decreased (p<0.05), while Parkin protein expression was significantly elevated in a dose-dependent manner (p<0.05). After treatment with Parkin in myocardial infarction rats, the relevant mRNA levels decreased, the number of apoptotic cells decreased, the myocardial fiber morphology returned to normal, the myocardial infarct size decreased, and the cardiac function of rats improved. Therefore, Parkin therapy plays an active role in cardiac function and ventricular remodeling in myocardial infarction rats.

Acute myocardial necrosis activates the immune response and inflammatory processes. Although the initial response is helpful in restoring tissue injury, dysregulated and exacerbated inflammation contributes to the progression of cardiac remodeling. Inflammasomes play important roles in post-infarction inflammation. NALP1/NLRP1, NLRP 3, and NLRC4 are the best-known inflammasomes. NLRP3, which has received the most study in cardiovascular disease, has been linked to increased IL-1β (IL1B) production and caspase-1 activity, as well as impaired cardiac function. The role of NLRP1 and NLRC4 inflammasomes after acute myocardial infarction (MI) is poorly understood. We evaluated the expression of myocardial inflammasomes and inflammatory markers 72 h after MI in rats. Male Wistar rats were divided into Sham (n = 15) and MI (n = 16) groups. MI was induced by ligating the left anterior descending coronary artery. Infarct size was assessed by histology. Myocardial protein and gene expression was analyzed by Western blot and RT-qPCR, respectively. IL-1β (Il1b) concentrations in serum and heart macerate supernatant were evaluated by ELISA. Statistical analysis was performed using Student's t test. Rats with an MI size less than 30% of the total left ventricle (LV) area were excluded; infarct size was 46 ± 11% of the total LV area in MI. The interstitial collagen fraction was higher in MI. Nlrc4, caspase-1 (Casp1), and IL-1β (Il1b) protein expressions were higher in MI. Nlrp3, Nlrp1, ASC (Pycard), pro-caspase-1, and pro-IL-1β (Il1b) expressions did not differ between groups. Expression of the Nlrp3 and ASC (Pycard) genes, as well as myocardial and serum IL-1β (Il1b) concentrations, was higher in MI. Acute post-myocardial infarction inflammation is characterized by increased protein expression of Nlrc4, caspase-1, and interleukin-1β; increased gene expression of Nlrp3 and ASC (Pycard); and elevated serum and myocardial concentrations of interleukin-1β in combination with an increased myocardial collagen interstitial fraction.

Triple nutrient supplementation improves survival, infarct size and cardiac function following myocardial infarction in rats

AimsThe aim of this study was to evaluate seven methods for quantifying myocardial oedema [2 standard deviation (SD), 3 SD, 5 SD, full width at half maximum (FWHM), Otsu method, manual thresholding, and manual contouring] from T2-weighted short tau inversion recovery (T2w STIR) and also to reassess these same seven methods for quantifying acute infarct size following ST-segment myocardial infarction (STEMI). This study focuses on test–retest repeatability while assessing inter- and intraobserver variability. T2w STIR and late gadolinium enhancement (LGE) are the most widely used cardiovascular magnetic resonance (CMR) techniques to image oedema and infarction, respectively. However, no consensus exists on the best quantification method to be used to analyse these images. This has potential important implications in the research setting where both myocardial oedema and infarct size are increasingly used and measured as surrogate endpoints in clinical trials.Methods and resultsForty patients day 2 following acute reperfused STEMI were scanned for myocardial oedema and infarction (LGE). All patients had a second CMR scan on the same day >6 h apart from the first one. Images were analysed offline by two independent observers using the semi-automated software. Both oedema and LGE were quantified using seven techniques (2 SD, 3 SD, 5 SD, Otsu, FWHM, manual threshold, and manual contouring). Interobserver, intraobserver and test–retest agreement and variability for both infarct size and oedema quantification were assessed. Infarct size and myocardial quantification vary depending on the quantification method used. Overall, manual contouring provided the lowest inter-, intraobserver, and interscan variability for both infarct size and oedema quantification. The FWHM method for infarct size quantification and the Otsu method for myocardial oedema quantification are acceptable alternatives.ConclusionsThis study determines that, in acute myocardial infarction (MI), manual contouring has the lowest overall variability for quantification of both myocardial oedema and MI when analysed by experienced observers.

Impact of low level laser irradiation on infarct size in the rat following myocardial infarction

The cellular and molecular mechanisms that underlie cardioprotection against I/R by anesthetic-induced preconditioning (APC) require further elucidation. Using isoflurane as a representative anesthetic, we evaluated the hypothesis that APC induces myocardial protection against I/R by attenuation of excessive reactive oxygen species and restoration of mitochondrial bioenergetics through postischemic up-regulation of manganese superoxide dismutase (MnSOD) expression and preservation of respiratory enzyme activity. Pentobarbital anesthetized open-chest Sprague-Dawley rats were subject to 30-min left coronary artery occlusion, followed by 120-min reperfusion. Before ischemia, rats were randomly assigned to receive 0.9% saline, two cycles of brief coronary artery occlusion and reperfusion, or a 30-min exposure to 1.0 minimum alveolar concentration isoflurane in the absence or presence of a specific mitochondrial adenosine triphosphate-sensitive potassium (KATP) channel blocker, 5-hydroxydecanoate; a membrane-permeable superoxide scavenger, 4-hydroxy-2,2,6,6-tetramethyl piperidinoxyl; or a NOS inhibitor, N(G)-nitro-L-arginine methyl ester. Isoflurane exposure induced an initial increase in myocardial superoxide (O2-), but not NO level. It also significantly decreased infarct size and restored mitochondrial respiratory enzyme activity or ATP production in I/R rat hearts, along with suppression of the O2- surge at reperfusion and increase in MnSOD expression or enzyme activity. These protective effects were abrogated by 5-hydroxydecanoate or 4-hydroxy-2,2,6,6-tetramethyl piperidinoxyl, but not by N(G)-nitro-L-arginine methyl ester pretreatment. These results suggest that opening of mitochondrial KATP channel, followed by O2- signaling, induces postischemic augmentation of MnSOD and preservation of mitochondrial respiratory enzyme activities, leading to attenuated cardiac O2- surge and restored ATP production during reperfusion, and underlie APC-induced cardioprotection.

Introduction Late gadolinium enhanced MRI (LGE-MRI) can quantify infarct size after myocardial infarction (MI) but is non-specific and reflects the increased membrane rupture and extracellular space that develop post MI.1 Mn is a potent T1-contrast agent that enters myocytes through active calcium channels, thus reducing T1 in viable myocardium.2 This active process rapidly ceases under ischemia. Hence, we hypothesised that Mn-enhanced MRI (MEMRI) could quantify final infarct size earlier than LGE-MRI. Methods Myocardial infarction was induced in 7 mice which then underwent MEMRI (n=4, 0.1 mmol/kg MnCl2) or LGE-MRI (n=3, 0.5 mmol/kg Gd-DTPA) at 1 hour post MI. All animals then underwent both MEMRI and LGEMRI at ∼24 hours post MI with a contrast washout period of at least 5 hours between scans. Imaging was performed using a 9.4T Agilent MRI system and a multi-slice inversion recovery sequence in the short-axis orientation covering the whole left ventricle TE/TR=3.04/1.11 ms, TI=∼600 ms for MEMRI and ∼350 ms for LGE-MRI, 900 excitation pulse, slice thickness=1.0 mm, FOV=25.6 × 25.6 mm, matrix size=128 × 128) as described.3 Results At 1 hour post MI, viable myocardium was enhanced in MEMRI, allowing early delineation of the occlusion zone as 41%±8% of the myocardium, whereas only subtle enhancement was seen on LGE-MRI, yielding a significantly lower value of 12%±2% (p=0.03. Figure 1). At ∼24 hours post MI, the MEMRI measure of infarct size remained constant (41%±5%) whilst LGE-MRI significantly increased to a level comparable with MEMRI (37%±3%). Figure 2 shows a direct comparison of MEMRI and LGE-MRI acquired in the same animal at 22 and 27 hours after MI, respectively, with matching histological TTC staining for infarct size. Discussion Effected myocytes rapidly stop internalising Mn under ischemic conditions, allowing early delineation of the occlusion zone. The membrane rupture that underlies LGE-MRI occurs later, meaning LGE-MRI underestimates the occlusion zone during the first hours post MI. Conclusions The present study shows MEMRI can quantify final infarct size earlier than LGE-MRI. This provides a sensitive approach which could be used as an early measure of cell death and myocardial viability when assessing the efficacy of new drugs which target acute MI. References . Doltra A, Amundsen BH, Gebker R, Fleck E, Kelle S. Emerging concepts for myocardial late gadolinium enhancement MRI. Current Cardiology Reviews2013;9:185. . Waghorn B, Schumacher A, Liu J, Jacobs S, Baba A, Matsuda T, Hu TC-C. Indirectly probing Ca(2+) handling alterations following myocardial infarction in a murine model using T(1)-mapping manganese-enhanced magnetic resonance imaging. Magnetic Resonance in Medicine2011;65:239. . Chow A, Stuckey DJ, Kidher E, Rocco M, Jabbour RJ, Mansfield CA, Darzi A, Harding SE, Stevens MM, Athanasiou T. Human induced pluripotent stem cell-derived cardiomyocyte encapsulating bioactive hydrogels improve rat heart function post myocardial infarction. Stem Cell Reports 2017;9:1415.

Ischemic heart disease is the leading cause of death in the United States, Canada, and worldwide. Severe disease is characterized by coronary artery occlusion, loss of blood flow to the myocardium, and necrosis of tissue, with subsequent remodeling of the heart wall, including fibrotic scarring. The current study aims to demonstrate the efficacy of quantitating infarct size via two-dimensional (2-D) echocardiographic akinetic length and four-dimensional (4-D) echocardiographic infarct volume and surface area as in vivo analysis techniques. We further describe and evaluate a new surface area strain analysis technique for estimating myocardial infarction (MI) size after ischemic injury. Experimental MI was induced in mice via left coronary artery ligation. Ejection fraction and infarct size were measured through 2-D and 4-D echocardiography. Infarct size established via histology was compared with ultrasound-based metrics via linear regression analysis. Two-dimensional echocardiographic akinetic length (r = 0.76, P = 0.03), 4-D echocardiographic infarct volume (r = 0.85, P = 0.008), and surface area (r = 0.90, P = 0.002) correlate well with histology. Although both 2-D and 4-D echocardiography were reliable measurement techniques to assess infarct, 4-D analysis is superior in assessing asymmetry of the left ventricle and the infarct. Strain analysis performed on 4-D data also provides additional infarct sizing techniques, which correlate with histology (surface strain: r = 0.94, P < 0.001, transmural thickness: r = 0.76, P = 0.001). Two-dimensional echocardiographic akinetic length, 4-D echocardiography ultrasound, and strain provide effective in vivo methods for measuring fibrotic scarring after MI.NEW & NOTEWORTHY Our study supports that both 2-D and 4-D echocardiographic analysis techniques are reliable in quantifying infarct size though 4-D ultrasound provides a more holistic image of LV function and structure, especially after myocardial infarction. Furthermore, 4-D strain analysis correctly identifies infarct size and regional LV dysfunction after MI. Therefore, these techniques can improve functional insight into the impact of pharmacological interventions on the pathophysiology of cardiac disease.