AI-Based Analysis of Tumor-Infiltrating Lymphocytes and Homologous Recombination in Ovarian Cancer: JGOG3025-A1 Study.

Understanding the tumor immune microenvironment, especially tumor-infiltrating lymphocytes (TILs), remains crucial in ovarian cancer. However, the distribution and prognostic significance of TILs across histological subtypes and genetic backgrounds remain unclear. As part of the JGOG3025-A1 study, diagnostic slides from 400 cases were collected. Two artificial intelligence-based cell classification models were used to evaluate the spatial distribution of TILs. The TIL score was calculated as the number of TILs divided by the analyzed area, and the immune-inflamed group was defined based on TIL scores. Among histological subtypes, high-grade serous carcinoma (HGSC) exhibited the highest TIL scores, clear cell carcinoma the lowest, and endometrioid carcinoma intermediate values. In HGSC, TIL scores did not significantly differ according to BRCA alteration or homologous recombination deficiency (HRD) status. The HRD/immune-inflamed group had the most favorable prognosis. In the homologous recombination-proficient population, the immune-inflamed group had better progression-free survival, whereas this trend did not appear for overall survival. In analyzes of the relationships between TIL levels and genomic structure, whole-genome doubling was associated with lower TIL infiltration in HRD tumors but showed no association in HRP tumors. In conclusion, HGSC showed the highest amount of TIL infiltration among subtypes, and prognostic stratification can be achieved by integrating pathology-based immunophenotypes and HRD status.