A deep learning pathomics model presented at AACR can stratify metastatic NSCLC patients by likely benefit from immune checkpoint inhibitors using routine pathology slides. Researchers from UT MD Anderson developed Path-IO and validated performance across multi-institution cohorts including Mayo Clinic, Gustave Roussy and the Phase III Lung-MAP S1400I trial, where high-risk groups showed more than double the risk of death or progression versus low-risk patients.