Quantifying Localized Tumor-Infiltrating Lymphocytes by Deep Learning on H&E Images: Analytical Validation and Prognostic Value in Lung Adenocarcinoma

The focal TIL score from H&E images is significantly associated with overall survival in a cohort of 685 ICI-treated lung adenocarcinoma patients (HR=0.60, P<0.001).

• Why it matters: Manual TIL scoring and methodological differences can obscure localized immune infiltration, limiting clinical utility and understanding of prognostic benefits for ICI-treated LUAD.
• What they did: Two deep learning models quantified lymphocyte density in tiles and defined a scorable tumor/tumor-adjacent stroma region, validated against IHC and consensus pathologist annotations (CCC=0.90; Dice=0.87).
• The result: In 685 ICI-treated LUAD patients, the focal TIL score was linked to better overall survival and outperformed PD-L1, supporting spatial intensity over global averages for immune response measurement.