In Cancer Research, AI Models Learn to See What Scientists Might Miss

Researchers compared multi-instance learning frameworks with attention mechanisms for classifying breast tumor whole slide images. Two approaches were evaluated for tumor and TP53 mutation detection in breast and lung cancers. Tumor detection was highly accurate (AUC >0.95), while mutation detection was more challenging (AUC <0.71). Higher image resolutions improved mutation detection. The study revealed opportunities to explore novel cancer morphological interpretations and could advance understanding of cancer etiology through interactive exploration of recurring tissue morphologies.