Prediction of mutations and outcome in gastrointestinal stromal tumors with deep learning: A multicenter, multinational study

Researchers analyzed 8,398 GIST cases using deep learning on whole-slide images to predict molecular mutations and patient prognosis. These models accurately identified KIT and PDGFRA variants while estimating sensitivity to treatments like imatinib. Additionally, the models predicted recurrence-free survival comparably to established pathology scores across international cohorts, providing a valuable tool for future clinical assessment.