New AI Could Make Cancer Treatment More Accurate

Researchers have developed an AI system that learns from medical data across many hospitals to avoid biases in cancer treatment. Most AI tools today are trained on limited data, which can reinforce stereotypes or miss rare cases. This new approach, called Mixture of Multicenter Experts (MoME), combines insights from different hospitals to create a more balanced and accurate model.

The AI focuses on prostate cancer radiation therapy, a treatment where doctors outline tumor targets on scans. Different hospitals often have different methods, leading to inconsistencies. This AI learns from multiple hospitals’ approaches without sharing private patient data, making it safer and more flexible. It even allows hospitals to adjust the AI to their local needs without exposing sensitive information.

Early tests show the AI performs better than existing tools, especially in cases with limited data or high variability between hospitals. This could mean more reliable treatments for patients, fewer errors, and less guesswork for doctors. The team behind the AI says it’s particularly useful for smaller or resource-limited hospitals that can’t afford expensive AI systems.

The big win here is privacy and fairness. Instead of pooling all patient data into one giant dataset—which raises privacy concerns—this AI learns from the patterns without seeing the actual records. That could make it easier for hospitals to adopt AI tools without risking patient confidentiality.