Abstract 2016: Prediction of prognosis in tumor-associated macrophage using immunohistochemistry with artificial intelligence

Abstract Tumor microenvironment (TME) is a cellular environment around the tumor. The quantification of lymphocyte populations called immunoscore has been developed and proven to have prognostic value. The impact of the degree of macrophage infiltration on the prognosis differs according to the type of cancer, and good prognosis with higher infiltration of macrophage is reported in colorectal cancer. Additionally, Image recognition technology based on deep learning is developing. There has been significant focus on classification by AI on immunohistochemistry (IHC) images. In this study, we evaluate correlation between tumor associated macrophage (TAM) distribution and prognosis of colorectal cancer, and we analyzed the data using artificial intelligence and deep learning techniques for predicting prognosis. Data from 159 patients who underwent R0 resection of primary colorectal cancer between April 2011 and December 2012 were reviewed. The 159 patients were divided into the following two cohorts: a training set of previous 100 patients and a validation set of 59 patients. Using the Anti-Iba-1 monoclonal antibody (NCNP27), we quantified TAM populations at the tumor invasive front in resected specimens and categorized by percentage of stained cytoplasm. A 0-75% density was assessed as low IM (invasive margin) score and a density between 75% and 100% was assessed as high IM score. In the training set, low IM score was also associated with poorer overall survival. The risk analysis of each factor for overall survival revealed significant difference in IM score. For the validation set, we employed MATLAB (Mathworks, R2021a) to analyze TAM distribution using a convolutional neural network (CNN). The sensitivity and specificity of the CNN mode for diagnosis were found to be 0.459 and 0.909, respectively. Consistent with the training results, a low IM score in the validation cohort was linked to the worst prognosis (P = 0.037). The findings emphasize the correlation between TAM distribution and patient prognosis in colorectal cancer, suggesting that IHC quantification of macrophages, when analyzed through AI techniques, offers valuable prognostic insights. This study highlights the potential of integrating AI with histopathological analysis to enhance prognostic prediction in oncology. Citation Format: Rie Mizumoto, Ayumi Nagae, Norikatsu Miyoshi, Kengo Haruna, Rie Hayashi, Mitsunobu Takeda, Yuki Sekido, Tsuyoshi Hata, Atsushi Hamabe, Takayuki Ogino, Mamoru Uemura, Yuichiro Doki, Hidetoshi Eguchi. Prediction of prognosis in tumor-associated macrophage using immunohistochemistry with artificial intelligence [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2016.