Artificial Intelligence-Powered Spatial Analysis of Tumor-Infiltrating Lymphocytes as Complementary Biomarker for...

Area of Science:

• Oncology
• Immunology
• Artificial Intelligence

Background:

• Tumor-infiltrating lymphocytes (TIL) show potential as biomarkers for immune checkpoint inhibitor (ICI) effectiveness.
• Current spatial analysis of TIL in whole-slide images (WSI) is limited and labor-intensive, hindering clinical application.

Purpose of the Study:

• To develop an artificial intelligence (AI)-powered WSI analyzer for spatial analysis of TIL.
• To define three immune phenotypes (IPs): inflamed, immune-excluded, and immune-desert.
• To correlate these IPs with tumor response to ICI and survival in advanced non-small-cell lung cancer (NSCLC).

Main Methods:

• Developed an AI-powered WSI analyzer to identify three immune phenotypes (inflamed, immune-excluded, immune-desert).
• Correlated identified IPs with tumor response and survival in two independent NSCLC cohorts.
• Validated AI-derived tumor proportion score (TPS) against pathologist-analyzed TPS.

Main Results:

• The inflamed IP correlated with higher response rates and prolonged progression-free survival (PFS) compared to immune-excluded or immune-desert phenotypes.
• Median PFS was 4.1 months for inflamed, 2.2 months for immune-excluded, and 2.4 months for immune-desert IPs.
• AI-derived TPS showed significant positive correlation with pathologist-analyzed TPS (P < .001).

Conclusions:

• AI-powered spatial analysis of TIL correlates with tumor response and PFS in advanced NSCLC patients treated with ICI.
• This AI tool offers a promising supplementary biomarker to pathologist-determined TPS for predicting ICI efficacy.
• The identified immune phenotypes can guide treatment decisions and prognostication in NSCLC.