Background: Automated interpretation of echocardiography by deep neural networks could support clinical reporting and improve efficiency. Whilst prior studies have evaluated spatial relationships using still frame images, our aim was to train and test a deep neural network for video analysis by combining spatial and temporal information, to automate the recognition of left ventricular regional wall motion abnormalities. Methods: We collected a series of transthoracic echocardiography examinations performed between July 2017 and April 2018 in two tertiary care hospitals. Regional wall abnormalities were defined by experienced physiologists and confirmed by trained cardiologists. First, we developed a 3-dimensional (3D) convolutional neural network (CNN) model for view selection ensuring stringent image quality control. Second, a U-net model segmented images to annotate the location of each left ventricular wall. Third, a final 3D CNN model evaluated echocardiographic videos from four standard views, before and after segmentation, and calculated a wall motion abnormality confidence level (0-1) for each segment. To evaluate model stability, we performed 5-fold cross-validation and external validation. Results: In a series of 10,638 echocardiograms, our view selection model identified 6,454 (61%) examinations with sufficient image quality in all standard views. In this training set, 2,740 frames were annotated to develop the segmentation model, which achieved a Dice similarity coefficient of 0.756. External validation was performed in 1,756 examinations from an independent hospital. A regional wall motion abnormality was observed in 8.9% and 4.9% in the training and external validation datasets, respectively. The final model recognized regional wall motion abnormalities in the cross-validation and external validation datasets with an area under the receiver operating characteristic curve of 0.912 (95% confidence interval [CI] 0.896 to 0.928) and 0.891 (95% CI 0.834 to 0.948), respectively. In the external validation dataset, the sensitivity was 81.8% (95% CI 73.8 to 88.2%) and specificity was 81.6% (95% CI 80.4 to 82.8%). Conclusions: In echocardiographic examinations of sufficient image quality, it is feasible for deep neural networks to automate the recognition of regional wall motion abnormalities using temporal and spatial information from moving images. Further investigation is required to optimise model performance and evaluate clinical applications.

Aim: Immunological checkpoint therapy is considered a powerful method for cancer therapy and acts by re-activating autologous T cells to kill the cancer cell. Myocarditis cases have been reported in cancer patients after immunological therapy; for example, nivolumab treatment is a monoclonal antibody that blocks programmed cell death-1/programmed cell death ligand-1 ligand interaction. This project provided insight into the inflammatory response as a benchmark to investigate the potential cardiotoxic effect of T cell response to the programmed cell death-1 (PD-1)/programmed cell death ligand-1 (PD-L1) axis in regulating cardiomyocyte injury in vitro. Methods and Results: We investigated cardiomyopathy resulted from the PD-1/PD-L1 axis blockade using the anti-PD-1 antibody in Rockefeller University embryonic stem cells-derived cardiomyocytes (RUES2-CMs) and a melanoma tumor-bearing murine model. We found that nivolumab alone did not induce inflammatory-related proteins, including PD-L1 expression, and did not induce apoptosis, which was contrary to doxorubicin, a cardiotoxic chemotherapy drug. However, nivolumab was able to exacerbate the immune response by increasing cytokine and inflammatory gene expression in RUES2-CMs when co-cultured with CD4⁺ T lymphocytes and induced apoptosis. This effect was not observed when RUES2-CMs were co-cultured with CD8⁺ T lymphocytes. The in vivo model showed that the heart function of tumor-bearing mice was decreased after treatment with anti-PD-1 antibody and demonstrated a dilated left ventricle histological examination. The dilated left ventricle was associated with an infiltration of CD4⁺ and CD8⁺ T lymphocytes into the myocardium. PD-L1 and inflammatory-associated gene expression were significantly increased in anti-PD-1-treated tumor-bearing mice. Cleaved caspase-3 and mouse plasma cardiac troponin I expressions were increased significantly. Conclusion: PD-L1 expression on cardiomyocytes suppressed T-cell function. Blockade of PD-1 by nivolumab enhanced cardiomyocyte inflammation and apoptosis through the enhancement of T-cell response towards cardiomyocytes.