Gao Peng, Xiao Qiong, Tan Hui, Song Jiangdian, Fu Yu, Xu Jingao, Zhao Junhua, Miao Yuan, Li Xiaoyan, Jing Yi, Feng Yingying, Wang Zitong, Zhang Yingjie, Yao Enbo, Xu Tongjia, Mei Jipeng, Chen Hanyu, Jiang Xue, Yang Yuchong, Wang Zhengyang, Gao Xianchun, Zheng Minwen, Zhang Liying, Jiang Min, Long Yuying, He Lijie, Sun Jinghua, Deng Yanhong, Wang Bin, Zhao Yan, Ba Yi, Wang Guan, Zhang Yong, Deng Ting, Shen Dinggang, Wang Zhenning
Department of Surgical Oncology and General Surgery, The First Hospital of China Medical University, Key Laboratory of Precision Diagnosis and Treatment of Gastrointestinal Tumors (China Medical University), Ministry of Education, Shenyang 110001, China.
School of Health Management, China Medical University, Shenyang 110122, China.
Cell Rep Med. 2024 Dec 17;5(12):101848. doi: 10.1016/j.xcrm.2024.101848. Epub 2024 Dec 4.
Neoadjuvant chemotherapy assessment is imperative for prognostication and clinical management of locally advanced gastric cancer. We propose an incremental supervised contrastive learning model (iSCLM), an interpretable artificial intelligence framework integrating pretreatment CT scans and H&E-stained biopsy images, for improved decision-making regarding neoadjuvant chemotherapy. We have constructed and tested iSCLM using retrospective data from 2,387 patients across 10 medical centers and evaluated its discriminative ability in a prospective cohort (132 patients; ChiCTR2300068917). iSCLM achieves areas under receiver operating characteristic curves of 0.846-0.876 across different test cohorts. Computed tomography (CT) and pathological attention heatmaps from Shapley additive explanations and global sort pooling illustrate additional benefits for capturing morphological features through supervised contrastive learning. Specifically, pathological top-ranked tiles exhibit decreased distances to tumor-invasive borders and increased inflammatory cell infiltration in responders compared with non-responders. Moreover, CD11c expression is elevated in responders. The developed interpretable model at the molecular pathology level accurately predicts chemotherapy efficacy.
新辅助化疗评估对于局部晚期胃癌的预后和临床管理至关重要。我们提出了一种增量监督对比学习模型（iSCLM），这是一个整合治疗前CT扫描和苏木精-伊红染色活检图像的可解释人工智能框架，用于改善新辅助化疗的决策。我们使用来自10个医疗中心的2387例患者的回顾性数据构建并测试了iSCLM，并在一个前瞻性队列（132例患者；ChiCTR2300068917）中评估了其判别能力。iSCLM在不同测试队列中的受试者操作特征曲线下面积达到0.846 - 0.876。来自Shapley加法解释和全局排序池化的计算机断层扫描（CT）和病理注意力热图说明了通过监督对比学习捕捉形态特征的额外益处。具体而言，与无反应者相比，病理排名靠前的切片在反应者中显示出与肿瘤浸润边界的距离减小以及炎症细胞浸润增加。此外，反应者中CD11c表达升高。在分子病理学水平上开发的可解释模型能够准确预测化疗疗效。
