Computational pathology assessments of cardiac stromal remodeling: Clinical correlates and prognostic implications in heart transplantation

Abstract

Background: The hostile immune environment created by allotransplantation can accelerate pathologic tissue remodeling. Both overt and indolent inflammatory insults propel this remodeling, but there is a paucity of tools for monitoring the speed and severity of remodeling over time. Methods: This retrospective cohort consisted of n = 2,167 digitized heart transplant biopsy slides along with records of prior inflammatory events and future allograft outcomes (cardiac death or allograft vasculopathy). Utilizing computational pathology analysis, biopsy images were analyzed to identify the pathologic stromal changes associated with future allograft loss or vasculopathy. Biopsy images were then analyzed to assess which historical inflammatory events drive progression of these pathologic stromal changes. Results: The top 5 features of pathologic stromal remodeling most associated with adverse allograft outcomes were also strongly associated with histories of both overt and indolent inflammatory events. Compared to controls, a history of high-grade or treated rejection was significantly associated with progressive pathologic remodeling and future adverse outcomes (32.9% vs 5.1%, p < 0.001). A history of recurrent low-grade rejection and Quilty lesions was also significantly associated with pathologic remodeling and adverse outcomes vs controls (12.7% vs 5.1%, p = 0.047). A history of high-grade or treated rejection in the absence of recurrent low-grade rejection history was not associated with pathologic remodeling or adverse outcomes (7.1% vs 5.1%, p = 0.67). Conclusions: A history of both traditionally treated and traditionally ignored alloimmune responses can predispose patients to pathologic allograft remodeling and adverse outcomes. Computational pathology analysis of allograft stroma yields translationally relevant biomarkers, identifying accelerated remodeling before adverse outcomes occur. Data availability: The data that support the findings of this study are presented in the manuscript and extended data sections. Unprocessed raw data are available from the corresponding author upon reasonable request. Source code for the stromal feature analysis pipeline is hosted on GitHub and freely available: https://github.service.emory.edu/CYUAN31/Pathomics_StromalBioMarker_in_Myocardium.git.