Heart transplant

AI system can help detect degree and severity of heart transplant rejection

Heart transplantation can be a life-saving operation for patients with end-stage heart failure. However, many patients experience organ transplant rejection, in which the immune system begins to attack the transplanted organ. But detecting transplant rejection is difficult -; in its early stages, patients may not experience symptoms, and experts disagree on the degree and severity of rejection.

To help address these challenges, researchers at Brigham and Women’s Hospital have created an artificial intelligence (AI) system known as the Cardiac Rejection Assessment Neural Estimator (CRANE) that can help detect rejection and to estimate its severity. In a pilot study, the team evaluated CRANE’s performance on samples provided by patients from three different countries, finding that it could help cardiac experts more accurately diagnose rejection and reduce the time needed for the exam. . The results are published in natural medicine.

Our retrospective pilot study demonstrated that the combination of artificial intelligence and human intelligence can improve expert agreement and reduce the time needed to assess biopsies. Our findings set the stage for large-scale clinical trials to establish the utility of AI models to improve heart transplant outcomes. »

Faisal Mahmood, PhD, lead author from the Mahmood Laboratory in the Department of Pathology at Brigham

Cardiac biopsies are commonly used to identify and assess the severity of organ rejection in patients after heart transplantation. However, several studies have shown that experts often disagree on whether the patient rejects the heart or how severe the rejection is. Diagnosis variability has direct clinical consequences, leading to treatment delays, unnecessary follow-up biopsies, anxiety, inadequate drug dosing, and ultimately poorer outcomes.

CRANE is designed to be used in tandem with expert assessment to establish an accurate diagnosis more quickly, and it can also be used in settings where there may be few pathology experts available. The team trained CRANE to detect, subtype and classify transplant rejection using thousands of pathological images from more than 1,300 Brigham heart biopsies. The researchers then validated the model, using test biopsies from the Brigham and independent external test sets received from hospitals in Switzerland and Turkey. The external validation datasets were constructed to demonstrate a high degree of variability to test the proposed AI model.

CRANE has achieved good results in the detection and assessment of rejection, with results comparable to those of conventional assessments. When experts used the tool, it reduced disagreements between experts and reduced evaluation time. The authors note that its use in clinical practice remains to be determined and plan to make further improvements to the system, but the results illustrate the potential for integrating AI into diagnostics.

“Throughout the history of medicine, diagnostic assessments have been largely subjective,” Mahmood said. “But because of the power and support of computer tools, that is starting to change. Now is the time to make a change by bringing together people with clinical expertise and those with computer science expertise to develop tools for assistive diagnosis.”


Brigham and Women’s Hospital

Journal reference:

Lipkova, J. et al. (2022) Deep learning-based assessment of cardiac allograft rejection from endomyocardial biopsies. Natural medicine. doi.org/10.1038/s41591-022-01709-2.