整合多通道2.5D Twins-SVT深度学习与影像组学的列线图预测局部进展期直肠癌原发TME术后无病生存
本研究回顾性纳入两家医院408例病理确诊的局部进展期直肠癌患者,均未接受新辅助放化疗而直接行原发全直肠系膜切除术。研究基于术前MRI提取影像组学特征并利用多通道2.5D Twins-SVT网络提取深度学习特征,结合临床因素构建预测无病生存的列线图。列线图在内部验证和外部测试中C-index分别为0.820和0.813,预测3年DFS的AUC分别为0.864和0.816,校准曲线和决策曲线分析显示良好预测效能和临床实用性。
要点
- 术前MRI影像组学联合2.5D深度学习列线图可对未行nCRT的LARC原发TME术后患者进行DFS风险分层。
- 列线图预测3年DFS的AUC在内部验证为0.864、外部测试为0.816,C-index分别为0.820和0.813。
- 决策曲线分析显示列线图在广泛临床决策阈值范围内具有临床实用性,可辅助个体化辅助治疗决策。
核心结论
- 整合多通道2.5D Twins-SVT深度学习与影像组学及临床因素的列线图能可靠预测未行nCRT的LARC患者原发TME术后的DFS,并支持个体化辅助治疗决策。
研究设计与样本
- 回顾性研究,纳入两家医院共408例病理确诊的LARC患者,均未接受nCRT而行原发TME。
- 患者分为训练队列189例、内部验证队列82例和外部测试队列137例。
- 从术前MRI提取影像组学特征,采用多通道2.5D Twins-SVT网络提取深度学习特征。
- 采用单因素和多因素Cox回归筛选预后显著临床因素,并与影像组学和2.5D深度学习特征整合构建列线图。
- 采用C-index、时间依赖AUC、校准曲线和决策曲线分析评价列线图预测性能。
诊断标准与关键数值
- 列线图预测DFS的C-index在内部验证为0.820,外部测试为0.813。
- 预测3年DFS的AUC在内部验证为0.864,外部测试为0.816。
- 校准曲线显示预测概率与实际结局接近一致。
- 决策曲线分析证实列线图在广泛临床决策阈值范围内提供显著临床效用。
临床应用与阅片要点
- 针对未接受nCRT而直接行原发TME的LARC患者,可基于术前MRI影像组学和深度学习特征联合临床因素使用列线图进行DFS风险分层。
- 列线图预测的3年DFS概率可辅助临床医生制定个体化辅助治疗决策。
局限性
- 本研究为回顾性设计。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVE: To develop a Nomogram for risk stratification of disease-free survival (DFS) in patients with locally advanced rectal cancer (LARC) undergoing primary total mesorectal excision (TME) without neoadjuvant chemoradiotherapy (nCRT), by integrating a multi-channel 2.5D Twins-SVT deep learning (DL) model and a radiomics model.
METHODS: In this retrospective study, a total of 408 patients with pathologically confirmed LARC from two hospitals were included. These patients were divided into a training cohort (n = 189), an internal validation cohort (n = 82), and an external test cohort (n = 137). Radiomic features were extracted from preoperative MRI images, and DL features were captured using a multi-channel 2.5D Twins-SVT network. Univariable and multivariable Cox regression analyses identified prognostically significant clinical factors, which were subsequently integrated with radiomic and 2.5D DL signatures into a predictive Nomogram. The predictive performance of this Nomogram was evaluated by employing the concordance index (C-index), time-dependent area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA).
RESULTS: The established Nomogram exhibited strong predictive performance regarding disease-free survival (DFS), reflected by C-index values of 0.820 in internal validation and 0.813 in the external testing cohort. When predicting 3-year DFS outcomes, the AUCs in the internal validation and external test cohorts reached 0.864 and 0.816, respectively. Calibration plots demonstrated close concordance between predicted probabilities and actual outcomes. Additionally, DCA confirmed that the Nomogram provided substantial clinical utility across a broad spectrum of clinical decision thresholds.
CONCLUSION: This Nomogram is a reliable tool to support clinicians in individualized adjuvant treatment decision-making for these patients.
引用
- Liu Z, Yang H, Nie L, et al. A nomogram integrating multi-channel 2.5D twins-SVT deep learning and radiomics for predicting disease-free survival in patients with locally advanced rectal cancer following primary total mesorectal excision. Eur J Radiol. 2026. PMID: 42691748. doi:10.1016/j.ejrad.2026.113196
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。