British journal of cancer · 2026
Deep learning-based prediction of lymph node metastasis and occult tumor cells in gastric cancer using histopathological images: a retrospective study.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
She Hairui, Xiang Tianqi, Wang Junjiang, Lin Zhihao, Huang Weiye, Wu Hongmei, Xu Yangwei, Wang Qian, Wu Xiaoxia, Chen Huihui, Chen Zhijie, Liang Li, Bian Xiuwu, Li Xiaomeng, Zhang Qingling
Аннотация
BACKGROUND: Lymph node metastasis (LNM) is an important prognostic factor but is often underdiagnosed due to limitations in conventional assessment methods. We aimed to develop a deep learning (DL) model to predict LNM status from primary gastric cancer (GC) whole-slide images.
METHODS: This retrospective study included 929 patients with GC from three independent cohorts across two centers. A DL model based on Clustering-constrained Attention Multiple Instance Learning was trained and validated to predict LNM from whole-slide images of primary tumors. The ability of the model to predict occult tumor cells (OTCs) in patients initially staged as pN0 and its prognostic value in patients with GC were examined.
RESULTS: The model demonstrated robust LNM predictive performance. Notably, it also predicted OTCs in patients initially staged as pN0. Patients staged as pN0 with OTCs had significantly worse survival outcomes than those staged as pN0 without OTCs (P = 0.003). The score generated by the model was an independent prognostic factor for patients with GC.
CONCLUSIONS: Our DL-based model enables accurate prediction of LNM and OTCs from primary GC slides, serving as a valuable tool for guiding personalized clinical strategies and as a novel biomarker for prognostic evaluation in patients with GC.
METHODS: This retrospective study included 929 patients with GC from three independent cohorts across two centers. A DL model based on Clustering-constrained Attention Multiple Instance Learning was trained and validated to predict LNM from whole-slide images of primary tumors. The ability of the model to predict occult tumor cells (OTCs) in patients initially staged as pN0 and its prognostic value in patients with GC were examined.
RESULTS: The model demonstrated robust LNM predictive performance. Notably, it also predicted OTCs in patients initially staged as pN0. Patients staged as pN0 with OTCs had significantly worse survival outcomes than those staged as pN0 without OTCs (P = 0.003). The score generated by the model was an independent prognostic factor for patients with GC.
CONCLUSIONS: Our DL-based model enables accurate prediction of LNM and OTCs from primary GC slides, serving as a valuable tool for guiding personalized clinical strategies and as a novel biomarker for prognostic evaluation in patients with GC.