AI-based quantitative histological analysis of non-metastatic regional lymph nodes predicts prognosis in MSI colorectal cancer

Regional lymph node metastasis is one of the main factors affecting solid cancer staging. However, the implications of non-metastatic regional lymph nodes (nrLNs) remain poorly understood. Here, we investigated the prognostic significance of the histological features of nrLNs in colorectal cancer (CRC) with microsatellite instability-high (MSI-H). Artificial intelligence-aided digital pathology-based quantification of 37 histological parameters in 873 whole-slide images comprising 5785 nrLNs was performed in two independent cohorts of MSI-H CRCs (discovery, n=103; validation, n=90). Germinal center (GC)-related factors, including the maximum area, mean area, sum area, and maximum diameter of GCs in the nrLNs, were identified as independent prognostic factors in MSI-H CRC. The prognostic GC-related features were significantly associated with peritumoral tertiary lymphoid structures and B cell pathways activation within nrLNs but were not or inversely correlated with tumor-infiltrating T cells and macrophages. In conclusion, quantitative GC-related histology of nrLNs may be a significant prognostic factor for cancer.