A recent study published by Co-Director of the Duke Division of Artificial Intelligence (AI) and Computational Pathology and Division Chief of Duke Cardiovascular Pathology Carolyn Glass, MD, PhD, et al. in Cardiovascular Pathology (Elsevier), highlighted the use of machine learning as an adjunct tool to diagnose difficult cases in heart transplant rejection.
Titled “A Machine Learning Algorithm Improves the Diagnostic Accuracy of the Histologic Component of Antibody Mediated Rejection (AMR-H) in Cardiac Transplant Endomyocardial Biopsies,” the study describes how a machine learning algorithm can reach acceptable performance of current diagnostic standards used to identify the histologic component (pAMR-H) necessary to provide an accurate diagnosis for heart transplant patients, and possibly surpass pathologist performance.
Even amongst the most experienced cardiovascular pathologists, certain diagnostic features of antibody-mediated rejection may be associated with higher interobserver variability. Identifying pAMR-H helps determine whether a patient needs to receive plasmapheresis, a method of removing blood plasma (which contains antibodies), from the body by withdrawing blood, separating it into plasma and cells, then transfusing the cells back into the bloodstream.
“For the first time, we determined if a machine learning algorithm could distinguish antibody mediated rejection from normal myocardium, healing injury, and acute cellular rejection,” said Glass. “I think the next real challenge is to see how this algorithm performs in real-life clinical scenarios outside the research setting at multiple high-volume institutions.”
The study represents a unique, synergistic collaboration between the Duke Department of Pathology, Duke Department of Biostatistics and Bioinformatics, and the Duke Pratt School of Engineering. The development and training phase of the novel deep-learning algorithm involved 16 expert cardiovascular pathologists across 12 different academic institutions. Read the study here.
Heart disease is the leading cause of death in the United States, and in 2023, the Duke Heart Transplant Center performed more heart transplants than any other center in the country. Duke’s Division of Cardiovascular Pathology provides comprehensive diagnostic expertise on cardiac and vascular disorders. Its cardiovascular pathologists include John Carney, MD, Louis DiBernardo, MD, Carolyn Glass, MD, PhD, Huihua Li, MD, and Elizabeth Pavlisko, MD, who support one of the highest-volume cardiac transplantation centers in the country and world. They evaluate some of the most complex and unique specimens within a tertiary academic setting. The mission of the team is to deliver the best possible patient care, to advance the field of cardiovascular pathology through scholarly activities at the national and international levels, and to train future leaders in the specialty.
Cardiovascular Pathology is the leading peer-reviewed journal dedicated to its field and is sponsored by the Society of Cardiovascular Pathology
