BMC medicine · 2026
Artificial intelligence empowers full-stack histopathological diagnosis and prognosis of renal cell tumor: a multi-center study with external validation.
СтатьяМногоцентровое исследованиеВалидационное исследованиеНазвание и аннотация: Перевод в очереди
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.
Xiong Ying, Xi Wei, Zhang Gelei, Luo Xiaoyuan, Xiao Li, Gao Jianbo, Wang Run, Wang Kang, Zhao Yun, Sun Qi, Wang Zilong, Guo Jianming, Qu Le, Hou Yingyong, Zhao Di, Wang Shuo
Аннотация
BACKGROUND: The rapid advancement of digital pathology has opened unprecedented opportunities for intelligent diagnosis in renal cell tumor. However, there remains a significant gap in the availability of reliable deep learning models capable of comprehensive kidney cancer detection, classification, grading, and survival prediction.
METHOD: This study retrospectively analyzed 11,135 whole-slide images (WSIs) from 7033 patients with renal tumor, sourced from four medical centers and two public cohorts. Histopathological representations were extracted using the foundation model Prov-GigaPath. A full-stack renal tumor diagnosis and prognosis framework was developed by combining fully supervised learning and weakly supervised multi-instance learning to enable both regional characterization and patient-level inference.
RESULTS: The deep learning model demonstrated high accuracy in identifying normal tissue (AUC = 0.990), tumor tissue (AUC = 0.982), necrosis tissue (AUC = 0.994), sarcomatoid differentiation (AUC = 0.967), and pseudocapsule tissue (AUC = 0.990) across various pathological types of renal cell tumor. For nine major subtypes of renal cell tumor, classification AUC reached 0.956-0.998 across multi-center validation cohorts. WHO/ISUP nuclear grade prediction for clear cell renal cell carcinoma (ccRCC) and papillary renal cell carcinoma (pRCC) achieved an AUC of 0.867. A whole-slide-derived pan-renal cell tumor pathological risk score independently predicted overall survival and significantly outperformed WHO/ISUP grading in prognostic stratification (p < 0.001).
CONCLUSIONS: We developed and validated a comprehensive AI framework integrating tissue-region detection, renal tumor subtype classification, nuclear grading, and survival prediction. These findings support its potential as a decision-support tool for renal tumor pathology, while prospective workflow-based studies are warranted to determine its clinical utility and impact on pathologist performance.
METHOD: This study retrospectively analyzed 11,135 whole-slide images (WSIs) from 7033 patients with renal tumor, sourced from four medical centers and two public cohorts. Histopathological representations were extracted using the foundation model Prov-GigaPath. A full-stack renal tumor diagnosis and prognosis framework was developed by combining fully supervised learning and weakly supervised multi-instance learning to enable both regional characterization and patient-level inference.
RESULTS: The deep learning model demonstrated high accuracy in identifying normal tissue (AUC = 0.990), tumor tissue (AUC = 0.982), necrosis tissue (AUC = 0.994), sarcomatoid differentiation (AUC = 0.967), and pseudocapsule tissue (AUC = 0.990) across various pathological types of renal cell tumor. For nine major subtypes of renal cell tumor, classification AUC reached 0.956-0.998 across multi-center validation cohorts. WHO/ISUP nuclear grade prediction for clear cell renal cell carcinoma (ccRCC) and papillary renal cell carcinoma (pRCC) achieved an AUC of 0.867. A whole-slide-derived pan-renal cell tumor pathological risk score independently predicted overall survival and significantly outperformed WHO/ISUP grading in prognostic stratification (p < 0.001).
CONCLUSIONS: We developed and validated a comprehensive AI framework integrating tissue-region detection, renal tumor subtype classification, nuclear grading, and survival prediction. These findings support its potential as a decision-support tool for renal tumor pathology, while prospective workflow-based studies are warranted to determine its clinical utility and impact on pathologist performance.