NIH funds validation of graft-versus-host disease assessment AI

Researchers from Vanderbilt University Medical Center (VUMC) have been awarded a five-year, $4.2 million grant from the National Heart, Lung, and Blood Institute, part of the National Institutes of Health (NIH), to validate an AI tool capable of measuring skin changes in chronic graft-versus-host disease (cGVHD) patients.

cGVHD is a potentially life-threatening complication following hematopoietic stem cell transplantation (HSCT).

HSCT, or bone marrow transplant, is a procedure that involves taking healthy hematopoietic stem cells from a donor and transferring them to a patient with bone marrow dysfunction. This infusion is valuable for the treatment of multiple conditions, including blood diseases and cancers.

cGVHD is a leading cause of morbidity and mortality for those who have undergone HSCT, making effective patient monitoring critical. Skin discoloration, inflammation, rash, itching and swelling are common in cGVHD, but current methods to assess these symptoms are both time-intensive and subjective.

To address this, the research team will work to validate an AI model to assess cGVHD patients’ skin changes.

“This grant will allow us to refine and validate an AI tool that can accurately and efficiently measure cGVHD skin changes from patient photographs. By improving the consistency, objectivity and efficiency of cGVHD assessments, we hope to enhance clinical trials and patient outcomes,” said Eric Tkaczyk, MD, PhD, assistant professor of Dermatology, Electrical and Computer Engineering, Biomedical Engineering and Biomedical Informatics at VUMC and director of the Vanderbilt Dermatology Translational Research Clinic, in the news release.

The researchers will develop a database made up of over 11,000 photographs and related clinical information from patient cohorts across five organizations:  Fred Hutchinson Cancer Center, Mayo Clinic, NIH, University of Pennsylvania and VUMC.

This data will be used to evaluate the AI, quantifying any potential biases that may exist in terms of photography conditions, disease severity, gender and skin tone. The tool’s accuracy will then be compared to that of expert dermatologists and in-person skin assessments.

Upon validation, the research team hopes that the model can advance at-home monitoring efforts, act as a resource for observational and therapeutic research, and reduce clinician burden.

“It is exhilarating to have the opportunity to validate the AI for clinical use after several years of work by the Vanderbilt team,” Tkaczyk noted.