KA-TMoE:基于时序CT放射组学预测NSCLC患者放疗后肋骨骨折的深度学习方法
本研究回顾性收集67例接受SBRT的NSCLC患者3个时间点CT,提出KA-TMoE模型整合纵向CT放射组学与放射组学分组知识,在肋骨水平预测骨折风险。验证队列AUC为0.792,模型预测高风险组骨折风险显著高于低风险组(HR=10.82,P<.001),多变量分析显示模型输出与骨折风险独立相关(OR=12.05,P=.002)。
要点
- KA-TMoE在验证队列预测SBRT后肋骨骨折的AUC为0.792。
- 模型预测高风险组骨折风险显著高于低风险组(HR=10.82,P<.001)。
- 模型输出在验证队列中与骨折风险独立相关(OR=12.05,P=.002)。
核心结论
- KA-TMoE作为初步的肋骨水平风险分层框架,在预测NSCLC患者SBRT后肋骨骨折方面显示出潜力。
研究设计与样本
- 回顾性研究收集67例NSCLC患者3个时间点的CT扫描,涵盖超过1600根肋骨。
- 提出KA-TMoE模型,整合纵向CT放射组学与放射组学分组知识,在肋骨水平估计骨折风险。
主要发现
- 验证队列中KA-TMoE的AUC为0.792。
- 时间-事件分析显示模型预测高风险组骨折风险高于低风险组(HR=10.82;P<.001)。
- 多变量logistic分析显示KA-TMoE输出在验证队列中与骨折风险独立相关(OR=12.05;P=.002)。
- 决策曲线分析显示在临床相关阈值范围内具有潜在净获益。
- 所有3个时间点的特征均对模型决策有贡献。
临床怎么用
- 模型可能支持更早的个体化风险分层、更密切的监测和及时的支持性评估。
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)。
原文
PURPOSE: Rib fracture is a recognized clinical complication in medically inoperable patients with non-small cell lung cancer (NSCLC) undergoing stereotactic body radiation therapy (SBRT), leading to diminished quality of life and delayed recovery. There remains an unmet need for reliable tools to predict rib-fracture risk to support individualized prognosis. This study aimed to develop and validate a deep learning model for predicting post-SBRT rib fractures using time-series computed tomography (CT) radiomics. METHODS AND MATERIALS: This retrospective study collected CT scans from 3 timepoints in 67 patients with NSCLC, comprising over 1600 individual ribs. We proposed a novel Knowledge-aware Temporal Mixture-of-Experts (KA-TMoE) model that integrates longitudinal CT radiomics with radiomic grouping knowledge to estimate fracture risk at the rib level. Model performance and interpretability were evaluated. RESULTS: The KA-TMoE model demonstrated favorable predictive performance in the validation cohort, achieving an area under the receiver operating characteristic curve (AUC) of 0.792. Exploratory DeLong testing was generally consistent with the observed performance differences between KA-TMoE and the ablation variants, suggesting that both longitudinal information and radiomics-grouping knowledge contributed to model performance. Mann-Whitney U tests demonstrated significant differences in model output distributions across cohorts. Time-to-event analysis showed that the model-predicted high-risk group had a higher risk of fracture than the low-risk group (hazard ratio = 10.82; P < .001). Multivariable logistic analysis showed that the KA-TMoE output remained independently associated with fracture risk in the validation cohort (odds ratio = 12.05; P = .002). Decision curve analysis demonstrated potential net benefit across clinically relevant thresholds. Features from all 3 timepoints contributed to the model's decision-making, highlighting the importance of temporal information. CONCLUSION: KA-TMoE showed potential as a preliminary rib-level risk-stratification framework for predicting post-SBRT rib fractures in patients with NSCLC. It may support earlier personalized risk stratification, closer surveillance, and timely supportive evaluation.
引用
- Chen Y, Farris M, Choi AR, et al. KA-TMoE: A Deep Learning Method Using Time-Series Computed Tomography Radiomics to Predict Rib Fractures after Radiation Therapy in Patients With NSCLC. Int J Radiat Oncol Biol Phys. 2026. PMID: 42575323. doi:10.1016/j.ijrobp.2026.07.056
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。