Computer-Assisted Histologic Evaluation of Cardiac Allograft Rejection

点击翻译按钮获取中文摘要

Project Summary Though cardiac transplantation is a lifesaving intervention, cardiac allograft rejection (CAR) remains a relatively common and serious complication that confers an increased risk of acute graft failure and adverse patient outcomes. For three decades, endomyocardial biopsy (EMB) with histological grading, as recommended by the International Society of Heart and Lung Transplantation (ISHLT) has been the broadly applied standard for CAR diagnosis. However, it is widely appreciated that the ISHLT rejection grading standard lacks diagnostic accuracy and has limited ability to discern the mechanism of rejection. These limitations expose patients to risks of both over-treatment and under-treatment, and highlight the unmet need for more accurate and informative approaches to histopathologic analysis of EMB samples. Our team is a leader in computational pathology image analysis with over 200 papers and >30 issued patents in this area. We have already developed and evaluated a computer assisted histopathology grading evaluation (CACHE) scheme which (1) in N=205 patients, had an area under the receiver operating characteristic curve (AUC)=0.95 compared to two cardiac pathologists (mean AUC=0.74) in distinguishing normal from failing hearts and (2) could distinguish low and high ISHLT rejection grades in N=1109 patients with a performance that exceeds that of trained cardiac pathologists. Recognizing the frequent discordance between ISHLT rejection grade and the clinical trajectory of a rejection event, we will further develop and optimize CACHE to identify new “grade agnostic” morphologic biomarkers of clinically serious CAR. Our scientific premise is that morphologic biomarkers prioritized based on their correlation to patients’ clinical trajectories and underlying immunological disease mechanisms will generate an accurate, consistent and informative classifier for diagnosing allograft rejection. In service of this hypothesis, the proposed research will address three specific aims. In Aim 1, we will utilize computational image analysis to discover the morphologic biomarkers of rejection-related injury which are needed to develop a classifier capable of assessing the clinical trajectory of CAR. In Aim 2, we will provide mechanistic annotation of biomarkers identified in Aim 1 through correlation with in-situ immunologic markers using custom multi-parameter immunofluorescence panels. In Aim 3, we employ a multicenter, prospective cohort to validate the diagnostic and mechanistic accuracy of the new rejection classifier developed in Aims 1 and 2. Ultimately, development of a more accurate and mechanistically informative tool for morphologic diagnosis of CAR will improve patient outcomes by reducing over- and under- treatment and inspire applications in other organ transplants. Interestingly, development of a superior histologic diagnostic tool will empower development of alternative, biopsy-free diagnostic approaches that have been handicapped by the necessity of comparison with the flawed ISHLT rejection grade as a reference standard.