Sai Kosaraju, Sai Phani Parsa, and Mingon Kang (all Computer Science) have published a research article titled, “Evidential deep learning-based ALK-expression screening using H&E-stained histopathological images," in npj Digital Medicine (Impact Factor: 15.1; ranked 2nd in Medical Informatics by Google Scholar Metrics). This publication summarizes a five-year international collaborative study conducted with Samsung Medical Center and Gyeongsang National University Hospital.

Identifying genetic alterations in non-small cell lung cancer (NSCLC) is a critical step in enabling targeted therapies. Lung cancer remains the leading cause of cancer-related deaths, responsible for over 127,000 deaths in 2023 — accounting for approximately 20% of all cancer fatalities. In particular, Anaplastic Lymphoma Kinase (ALK) rearrangement represents one of the most important therapeutic targets, leading to substantial improvements in survival among patients with advanced NSCLC.

Our study demonstrates over 95% predictive accuracy with ALK-specific morphological evidence for trustworthy predictions using diverse datasets from both resection and biopsy samples, showing strong potential for clinical application. Several unique features distinguish this work:

• Novel methodology: We introduce an advanced deep learning framework, DeepPATHO, that analyzes complex pathological morphologies across multiple magnification levels, substantially enhancing predictive performance.

• Extensive dataset: The study utilizes the largest collection of ALK-positive whole-slide images, accumulated over ten years from one of South Korea’s largest medical centers.

• Clinical validation: The proposed model was validated across multiple datasets, and its pathological evidence was rigorously reviewed in consultation with several pathologists.