AI-Powered Surveillance of Immunotherapy Skin Toxicities in Melanoma

The Problem: New immune therapies have dramatically improved survival for melanoma patients, but they can sometimes make the immune system attack healthy skin, causing rashes and other side effects. These skin problems can be painful, disruptive, and even force patients to stop life-saving treatment. Studying them is very slow and hard because doctors would have to read through thousands of patient records by hand. Our Solution: We built a “smart AI team” that can read patient records like a group of expert assistants. Each AI has a job: one identifies the type of rash, another finds when it started, a third checks whether the therapy likely caused it, and a fourth grades how severe it is. A supervising AI double-checks all their work. The AI organizes the results into a standardized digital format, like putting together furniture by following a specific blueprint, so they can be used safely and consistently across hospitals. What We Did: We first tested the AI on 2,000 patient charts reviewed by real doctors to see how well it matches human judgment. We also tested it on 700 charts from another hospital to make sure it works in different settings. The system is designed to avoid mistakes, and to give reliable results every time. The Impact: This AI team can work more than 50 times faster than humans and still be accurate. It will create a large, clean dataset that links skin side effects to patient genetics, helping researchers discover who is most at risk. Ultimately, this could guide safer, more personalized treatment for melanoma patients.