Evidential deep learning-based ALK-expression screening using H&E-stained histopathological images

1

Deep learning can accurately predict ALK rearrangements in non-small cell lung cancer from H&E-stained images, improving diagnostic efficiency.
2

The proposed DeepPATHO model achieved over 95% accuracy in identifying ALK alterations from both resection and biopsy datasets.
3

Current ALK screening methods are inefficient, leading to unnecessary costs for the majority of NSCLC patients who do not have ALK rearrangements.
4

Limited tissue availability and the challenges of accessing lesions complicate the diagnosis of ALK positivity in advanced lung cancer stages.
5

A publicly available Python-based software package supports the clinical application of the developed deep learning model for ALK screening.

Npj Digital Medicine

by Sai Chandra Kosaraju, Sai Phani Parsa, Dae Hyun Song, Hyo Jung An, Yoon-La Choi, Joungho Han, Jung Wook Yang, Mingon Kang
October 14, 2025

1