Interpretable deep learning maps spatial morphological phenotypes to HCC and LUAD survival.

MorphX, an interpretable deep-learning framework, predicts survival in hepatocellular carcinoma (HCC) and lung adenocarcinoma (LUAD) with concordance indices of 0.775–0.799 in discovery cohorts.

• Why it matters: Black-box pathology models limit clinical trust because they do not clearly link prognosis to biological morphology.
• What they did: In a multi-center study of 1,067 patients, MorphX decodes multi-scale patterns from whole-slide images by combining macro tissue texture with micro nuclear metrics, then derives eight morphology-based biomarkers tied to features like necrotic texture heterogeneity, lymphocyte density, and nuclear shape.
• The result: The eight biomarkers show reproducible survival associations across independent cohorts, enabling more granular, biology-linked estimates of tumor aggressiveness beyond traditional histological grading.

AI summary of the abstract · may contain errors · not medical advice · read the paper.