基于融合影像组学预测骨肉瘤对新辅助化疗的反应
本研究回顾性分析106例病理确诊骨肉瘤患者的轴位T2WI和增强T1WI,构建深度学习影像组学模型预测新辅助化疗反应。自动分割Dice系数0.868,模型在测试集准确率93.8%、AUC 0.961,可识别化疗反应差的患者。
要点
- 自动分割框架对骨肉瘤的Dice系数达0.868。
- DLR模型预测NAC反应测试集准确率93.8%、AUC 0.961。
- 模型可识别NAC反应差者,辅助风险咨询和治疗决策。
核心结论
- DLR模型可作为实用预测工具,识别NAC反应差的患者,为风险咨询和临床治疗决策提供信息;反应差者建议接受免疫治疗和最佳支持治疗。
研究设计与样本
- 回顾性研究,纳入106例病理确诊骨肉瘤患者,使用轴位T2WI和增强T1WI,采用Auto3DSeg框架自动分割,基于三种特征提取方法和三种分类器构建九种风险分类模型,并开发含临床指标的深度学习影像组学列线图。
诊断标准与关键数值
- 自动分割模型跨数据集Dice系数0.868;DLR模型预测NAC反应在测试集准确率93.8%、AUC 0.961;计算了AUC、敏感度、特异度、准确率、阴性预测值和阳性预测值进行性能评估,并用校准曲线和决策曲线分析评估预测能力和临床净获益。
临床应用与阅片要点
- 作者认为DLR模型可作为实用预测工具,识别NAC反应差的患者,为风险咨询和临床治疗决策提供信息;反应差者建议接受免疫治疗和最佳支持治疗。
局限性
- 原文未报告局限性。
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
RATIONALE AND OBJECTIVES: Neoadjuvant chemotherapy (NAC) is the most crucial prognostic factor for osteosarcoma (OS), it significantly prolongs progression-free survival and improves the quality of life. This study aims to develop a deep learning radiomics (DLR) model to accurately predict the response to NAC in patients diagnosed with OS using preoperative MR images.
METHODS: We reviewed axial T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted (T1CE) of 106 patients pathologically confirmed as OS. First, the Auto3DSeg framework was utilized for automated OS segmentation. Second, using three feature extraction methods, nine risk classification models were constructed based on three classifiers. The area under the receiver operating curve (AUC), sensitivity, specificity, accuracy, negative predictive value and positive predictive value were calculated for performance evaluation. Additionally, we developed a deep learning radiomics nomogram with clinical indicators.
RESULTS: The model for OS automatic segmentation achieved a Dice coefficient of 0.868 across datasets. To predict the response to NAC, the DLR model achieved the highest prediction performance with an accuracy of 93.8% and an AUC of 0.961 in the test sets. We used calibration curves to assess the predictive ability of the models and performed decision curve analysis to evaluate the clinical net benefit of the DLR model.
CONCLUSION: The DLR model can serve as a pragmatic prediction tool, capable of identifying patients with poor response to NAC, providing information for risk counseling, and assisting in making clinical treatment decisions. Poor responders are better advised to undergo immunotherapy and receive the best supportive care.
引用
- Zheng F, Yin P, Liang K, et al. Fusion Radiomics-Based Prediction of Response to Neoadjuvant Chemotherapy for Osteosarcoma. Acad Radiol. 2023. PMID: 38151381. doi:10.1016/j.acra.2023.12.015
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。