A computed tomography-based multitask deep learning model for predicting tumour stroma ratio and treatment outcomes in patients with colorectal cancer: a multicentre cohort study

作者全名：Cui, Yanfen; Zhao, Ke; Meng, Xiaochun; Mao, Yun; Han, Chu; Shi, Zhenwei; Yang, Xiaotang; Tong, Tong; Wu, Lei; Liu, Zaiyi

作者地址：[Cui, Yanfen; Zhao, Ke; Han, Chu; Shi, Zhenwei; Wu, Lei; Liu, Zaiyi] Southern Med Univ, Guangdong Prov Peoples Hosp, Guangdong Acad Med Sci, Dept Pharm, Guangzhou, Peoples R China; [Cui, Yanfen; Zhao, Ke; Han, Chu; Shi, Zhenwei; Wu, Lei; Liu, Zaiyi] Guangdong Prov Key Lab Artificial Intelligence Med, Guangzhou, Peoples R China; [Cui, Yanfen; Zhao, Ke; Han, Chu; Shi, Zhenwei; Wu, Lei; Liu, Zaiyi] Guangdong Acad Med Sci, Guangdong Prov Peoples Hosp, Guangdong Cardiovasc Inst, Guangzhou, Peoples R China; [Meng, Xiaochun] Sun Yat sen Univ, Affiliated Hosp 6, Dept Radiol, Guangzhou, Peoples R China; [Cui, Yanfen; Yang, Xiaotang] Shanxi Med Univ, Chinese Acad Med Sci, Shanxi Prov Canc Hosp, Canc Hosp,Shanxi Hosp,Canc Hosp,Dept Radiol, Taiyuan 030013, Peoples R China; [Mao, Yun] Chongqing Med Univ, Affiliated Hosp 1, Dept Radiol, Chongqing, Peoples R China; [Tong, Tong] Fudan Univ, Shanghai Canc Ctr, Dept Radiol, Shanghai 200032, Peoples R China

通信作者：Wu, L; Liu, ZY (通讯作者)，Southern Med Univ, Guangdong Prov Peoples Hosp, Guangdong Acad Med Sci, Dept Pharm, Guangzhou, Peoples R China.; Yang, XT (通讯作者)，Shanxi Med Univ, Chinese Acad Med Sci, Shanxi Prov Canc Hosp, Canc Hosp,Shanxi Hosp,Canc Hosp,Dept Radiol, Taiyuan 030013, Peoples R China.; Tong, T (通讯作者)，Fudan Univ, Shanghai Canc Ctr, Dept Radiol, Shanghai 200032, Peoples R China.

来源：INTERNATIONAL JOURNAL OF SURGERY

ESI学科分类：CLINICAL MEDICINE

WOS号：WOS:001223156300018

JCR分区：Q1

影响因子：12.5

年份：2024

卷号：110

期号：5

开始页：2845

结束页：2854

文献类型：Article

关键词：Colorectal cancer; deep learning; survival; tumour-stroma ratio

摘要：Background: Tumour-stroma interactions, as indicated by tumour-stroma ratio (TSR), offer valuable prognostic stratification information. Current histological assessment of TSR is limited by tissue accessibility and spatial heterogeneity. The authors aimed to develop a multitask deep learning (MDL) model to noninvasively predict TSR and prognosis in colorectal cancer (CRC). Materials and methods: In this retrospective study including 2268 patients with resected CRC recruited from four centres, the authors developed an MDL model using preoperative computed tomography (CT) images for the simultaneous prediction of TSR and overall survival. Patients in the training cohort (n=956) and internal validation cohort (IVC, n=240) were randomly selected from centre I. Patients in the external validation cohort 1 (EVC1, n=509), EVC2 (n=203), and EVC3 (n=360) were recruited from other three centres. Model performance was evaluated with respect to discrimination and calibration. Furthermore, the authors evaluated whether the model could predict the benefit from adjuvant chemotherapy. Results: The MDL model demonstrated strong TSR discrimination, yielding areas under the receiver operating curves (AUCs) of 0.855 (95% CI, 0.800-0.910), 0.838 (95% CI, 0.802-0.874), and 0.857 (95% CI, 0.804-0.909) in the three validation cohorts, respectively. The MDL model was also able to predict overall survival and disease-free survival across all cohorts. In multivariable Cox analysis, the MDL score (MDLS) remained an independent prognostic factor after adjusting for clinicopathological variables (all P<0.05). For stage II and stage III disease, patients with a high MDLS benefited from adjuvant chemotherapy [hazard ratio (HR) 0.391 (95% CI, 0.230-0.666), P=0.0003; HR=0.467 (95% CI, 0.331-0.659), P<0.0001, respectively], whereas those with a low MDLS did not. Conclusion: The multitask DL model based on preoperative CT images effectively predicted TSR status and survival in CRC patients, offering valuable guidance for personalized treatment. Prospective studies are needed to confirm its potential to select patients who might benefit from chemotherapy.

基金机构：Key-Area Research and Development Program of Guangdong Province, China [2021B0101420006]; Regional Innovation and Development Joint Fund of National Natural Science Foundation of China [U22A20345]; Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application [2022B1212010011]; National Science Fund for Distinguished Young Scholars of China [81925023]; National Natural Science Foundation of China [82371952, 82271946, 82202267, 82171923, 82102019, 82001789]; High-level Hospital Construction Project [DFJHBF202105]; Applied Basic Research Projects of Shanxi Province, China, Outstanding Youth Foundation [202103021222014]

基金资助正文：This study was supported by the Key-Area Research and Development Program of Guangdong Province, China (No.2021B0101420006),Regional Innovation and Development Joint Fund of National Natural Science Foundation of China (No. U22A20345), Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application (No. 2022B1212010011), National Science Fund for Distinguished Young Scholars of China (No. 81925023), National Natural Science Foundation of China (No. 82001789, 82102019, 82171923, 82202267, 82271946, 82371952), High-level Hospital Construction Project (No. DFJHBF202105), and the Applied Basic Research Projects of Shanxi Province, China, Outstanding Youth Foundation (No. 202103021222014).