An NYU Langone Health study published on September 25, 2026, shows that combining electrocardiograms (EKGs) with blood tests analyzed by artificial intelligence can accurately detect heart transplant rejection. In a test of 38 patients, the combined AI model correctly identified 94% of those not experiencing rejection, potentially sparing them from invasive surgical biopsies. Researchers plan to validate the model in multi-center trials.
Significance and future of the study
- ▪As a next step, the NYU Langone Health researchers plan to test their combined artificial intelligence model in more patients at several transplant centers.
- ▪A study published in JHLT Open on September 25, 2026, led by NYU Langone Health researchers, is the first to combine blood biomarkers with electrocardiogram readings in an artificial intelligence model and compare the analysis against biopsy results.
Training the artificial intelligence models
- ▪One artificial intelligence model was trained on electrocardiogram data alone, while another model combined electrocardiogram readings with the results of two blood tests commonly used to predict heart transplant rejection risk.
- ▪The NYU Langone Health researchers matched each electrocardiogram to a biopsy performed within the past month and grouped organ rejections into no or mild rejection versus moderate or severe rejection.
- ▪NYU Langone Health researchers trained artificial intelligence models on 5,300 electrocardiogram readings taken from 2,357 adult heart transplant recipients treated between 2018 and 2024.
Performance of the AI models
- ▪In a test group of 38 heart transplant recipients, the artificial intelligence model based on blood tests alone incorrectly flagged 19 patients as potentially needing a biopsy.
- ▪In a test group of 38 heart transplant recipients, the combined artificial intelligence model correctly identified 94% of patients who were not experiencing rejection, potentially sparing them from unnecessary biopsies.
Senior author's commentary
- ▪The study's senior author is Lior Jankelson, M.D., Ph.D., an associate professor at NYU Grossman School of Medicine and NYU Tandon School of Engineering.
- ▪Lior Jankelson stated that combining electrocardiograms with artificial intelligence could help identify heart transplant rejection earlier and save more lives.
Current heart transplant rejection tests
- ▪Although effective at spotting heart transplant rejection, blood biomarkers frequently produce false-positive test results that lead to unnecessary surgical biopsies.
- ▪Blood biomarkers used to predict heart transplant rejection measure gene activity linked to cellular rejection and fragments of donor DNA in the recipient's bloodstream.
- ▪The current gold standard for diagnosing heart transplant rejection is a surgical biopsy, which involves removing a small piece of heart muscle to inspect for immune system rejection.
Debatable claims
- ▪Transplant centers should delay adopting AI diagnostic models until multi-center trials are complete
- ▪AI models are reliable enough to guide critical heart transplant rejection decisions
- ▪Surgical biopsies remain necessary for diagnosing heart transplant rejection despite AI advancements
Story comments
Loading comments…