An Interpretable Deep Learning Model for Predicting Microsatellite State in Colorectal Cancer
*Corresponding Author: Zhenqi Zhang, Department of Pathology, The Affiliated Hospital Of Qingdao University, Qingdao, ChinaReceived Date: Dec 12, 2024 / Published Date: Sep 17, 2026
Citation: Zhang Z, Wang W, Song Y, Liu X, Yang P, et al. (2026) An Interpretable Deep Learning Model for Predicting Microsatellite State in Colorectal Cancer. Diagnos Pathol Open 11: 276.
Copyright: Zhang Z, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution and reproduction in any medium, provided the original author and source are credited.
Abstract
Microsatellite (MS) state detection is crucial for precise diagnosis and treatment of Colorectal Cancer (CRC). However, limited medical resources and economic challenges make MS state testing inaccessible for many CRC patients. We propose a novel pathology-interpretable deep learning model to predict the MS state of CRC, validated across multiple cohorts and sequencing platforms, especially in the Asian population. We included pathological images, documented MS state and clinical characteristics of 360 CRC patients from The Cancer Genome Atlas and 782 cases from Chinese hospitals. The model demonstrated notable generalization ability, achieving an AUC of 0.92 in an independent verification cohort and 0.93 in a multicenter cohort. Using the Hover-Net model, we achieved cell nucleus segmentation and image-based cell type identification, providing pathology interpretation of the model and identifying significant predictors of microsatellite instability. Patients classified as Microsatellite Instability (MSI) by the model showed higher progression-free survival rates, supported by follow-up data. The model shows great potential for clinical usage, particularly in the Asian population, demonstrating high accuracy across multiple cohorts and MSI detection platforms.
