Objective:

To develop an interpretable prediction model for lung adenocarcinoma immune subtyping by quantifying spatial distribution patterns of tumor-infiltrating lymphocytes in H&E whole-slide images, providing a computational tool for tumor immune microenvironment evaluation.

Key Findings:

• 503 LUAD patients stratified into high- and low-immunity subgroups based on transcriptomic analysis.
• High-immunity group showed elevated CD8+ T cells, M1 macrophages, and higher tumor mutation burden.
• Automated annotation model achieved high accuracy in tissue segmentation and TIL identification, with a 79.51% F1-score.
• Immune subtype classification model achieved AUC of 0.839 in internal and 0.927 in external validation cohorts.

Interpretation:

The model links morphological phenotypes to molecular immune subtypes, providing a transparent and verifiable tool for assessing tumor immune status, with potential implications for clinical decision-making.

Limitations:

• High cost of transcriptome sequencing limits routine clinical application, impacting accessibility.
• Current evaluation relies on subjective visual assessments by pathologists, affecting consistency.

Conclusion:

The study presents a cost-effective and scalable tool for immune subtype prediction in LUAD, aiding in immunotherapy decision-making.