Eliot Peyster, MD, MSc
Instructor, Advanced Heart Failure and Transplant Medicine
University of Pennsylvania
Traditional histology remains a vital part of post-heart transplant care, with recipients undergoing frequent surveillance biopsies for histologic rejection grading. Unfortunately, rejection grading suffers from poor inter-pathologist agreement, poor correlation with the clinical rejection syndrome the patient experiences, and offers no prognostic value to guide future biopsy schedules or immunosuppression decisions. These limitations create uncertainty in clinical care, present a barrier to multicenter research, and expose patients to the risks of both over- and under-treatment. Utilizing advances in computer processing power and machine learning methods, it is now possible to perform rigorous and comprehensive feature quantification of medical image data. While a recognized emerging technology within oncologic medicine, computerized analysis of histologic samples has thus far had minimal impact within cardiovascular medicine. In this presentation, I will review the limitations of the current diagnostic standard for diagnosing allograft rejection, introduce the basic methodology underlying computational histologic analysis, and describe several novel applications demonstrating the translational potential of this technology in the era of precision medicine.