Abstract 276: Artificial intelligence-powered analysis of tumor-stroma ratio and fibroblast density as prognostic indicators in colorectal cancer

Abstract

Abstract Introduction: The tumor microenvironment (TME) plays a critical role in cancer prognosis. In colorectal cancer (CRC), the tumor-stroma ratio (TSR) and cancer-associated fibroblast (CAF) density are emerging as prognostic markers. However, accurately quantifying them remains challenging. Recent studies have shown that artificial intelligence (AI)-powered analyzers can quantify features in the TME including these. Methods: Lunit SCOPE IO, an AI-powered H&E whole-slide image analyzer that detects fibroblasts and segments cancer area and stromal area, was applied to evaluate TSR and CAF density in treatment-naive, stage II or III CRC surgical specimens (N = 207) from Severance Hospital. TSR was calculated as the ratio of the area of the cancer area (CA) to the sum of the CA and the cancer stromal area (CS) within a slide. CAF density was evaluated as the fibroblast density within the CA or CS, or the ratio between them. Cox survival analysis determined the hazard ratios (HRs) in progression-free survival (PFS) between the high/low TSR/CAF groups, defining high/low groups based on optimized hazard ratio (HR) cutoffs. Results: The study found a mean TSR of 0.481±0.126 and CAF densities of 69.96±37.53/mm² (cancer area) and 1844.29±440.31/mm² (stromal area). Higher TSR, lower CAF density in CA, and higher CAF density in CS were linked to better prognosis (Table 1). The mean CAF ratio (CAF density of CS divided by CAF density of CA) was 33.92±25.35, and a group with high values showed a favorable prognosis. Dividing patients based on TSR and CAF ratio, we found that high levels in either or both metrics predicted better outcomes compared to low levels in both. Table 1. Prognosis by tumor-stroma ratio (TSR) and cancer-associated fibroblast (CAF) density Factor Group Hazard ratio (95% confidence interval) Mean progression-free survival (months) p-value TSR (cutoff: 0.351) Low (N = 34) Reference 43.7 <0.001 High (N = 173) 0.32 (0.19 - 0.56) 67.2 CAF density of cancer area (CA) (cutoff: 46.8) Low (N = 62) Reference 69.3 0.056 High (N = 145) 1.86 (0.98 - 3.50) 60.7 CAF density of cancer stromal area (CS) (cutoff:1,901) Low (N = 116) Reference 58.2 0.015 High (N = 91) 0.50 (0.29 - 0.88) 69.9 The ratio of CAF density of CS to CA (CAF ratio) (cutoff: 53.1) Low (N = 187) Reference 61.5 0.043 High (N = 20) 0.13 (0.02 - 0.94) 80.9 Combination of TSR and CAF ratio Both low (N = 32) Reference 41.2 <0.001 TSR high, CAF ratio low (N = 155) 0.32 (0.19 - 0.56) 65.8 TSR low, CAF ratio high (N = 2) 0.00 (0.00 - ∞) 83.4 Both high (N = 18) 0.06 (0.01 - 0.45) 80.2 Conclusions: In this study, we found a favorable prognosis in the group with higher TSR or higher density of CAF in CS compared to CA. This result suggests that AI-enhanced analysis of TSR and CAF can effectively predict prognosis in CRC, highlighting the potential roles of these markers in cancer prognosis. Citation Format: Minsun Jung, Jun Yong Kim, Mingu Kang, Jinhee Lee, Sanghoon Song, Taebum Lee, Wonkyung Jung, Soo Ick Cho. Artificial intelligence-powered analysis of tumor-stroma ratio and fibroblast density as prognostic indicators in colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 276.