基于多区域超声影像组学特征联合病理参数预测乳腺癌新辅助化疗疗效的分析
本研究回顾性纳入2019年1月至2024年12月在兰州大学第二医院接受新辅助化疗的321例乳腺癌患者,从二维超声图像勾画瘤内及瘤周5、10、15 mm区域并提取影像组学特征,构建随机森林模型预测病理完全缓解。瘤周10 mm联合瘤内模型在验证集AUC为0.842,联合HER-2、Ki-67后AUC达0.869,DeLong检验显示联合模型显著优于其他模型。
要点
- 瘤周10 mm是预测新辅助化疗后病理完全缓解的最佳瘤周区域。
- 瘤内联合瘤周10 mm模型验证集AUC为0.842(95%CI 0.764-0.921)。
- 联合HER-2、Ki-67后模型AUC升至0.869(95%CI 0.800-0.937)。
核心结论
- 瘤周10 mm区域是乳腺癌新辅助化疗后肿瘤周围组织的最佳预测区域,纳入临床病理参数(HER-2、Ki-67)的IntraPeri-10 mm联合模型预测新辅助化疗疗效表现更优,可准确评估治疗反应,为后续治疗决策提供有价值指导。
研究设计与样本
- 回顾性研究纳入2019年1月至2024年12月在兰州大学第二医院接受新辅助化疗的321例乳腺癌患者,按术后病理分为病理完全缓解组与非病理完全缓解组,按7:3随机分为训练集224例和验证集97例,使用Itk-snap软件在二维超声图像上勾画瘤内及瘤周5、10、15 mm区域并提取影像组学特征,采用随机森林机器学习分类器建模,以受试者工作特征曲线下面积评估预测性能。
诊断标准与关键数值
- 共提取214个影像组学特征,经最小绝对收缩和选择算子回归筛选出与病理完全缓解密切相关的瘤内特征8个、瘤周10 mm特征8个、瘤内联合瘤周10 mm特征9个。瘤内模型训练集与验证集AUC分别为0.860和0.823;瘤周5 mm、10 mm、15 mm模型训练集AUC分别为0.836、0.854、0.822,验证集分别为0.793、0.799、0.792;验证集中瘤内联合瘤周10 mm模型AUC为0.842(95%CI 0.764-0.921),优于瘤内联合瘤周5 mm模型(AUC 0.831,95%CI 0.758-0.914)和瘤内联合瘤周15 mm模型(AUC 0.838,95%CI 0.761-0.917);基于瘤内联合瘤周10 mm与临床病理参数(HER-2、Ki-67)的联合模型AUC为0.869(95%CI 0.800-0.937),DeLong检验显示联合模型AUC显著优于其他模型。
临床应用与阅片要点
- 作者认为瘤周10 mm区域是乳腺癌新辅助化疗后肿瘤周围组织的最佳预测区域,纳入临床病理参数的IntraPeri-10 mm模型预测新辅助化疗疗效表现更好,可准确评估治疗反应,为后续治疗决策提供有价值指导。
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
OBJECTIVE: This study aimed to analyze the correlation between the ultrasonographic radiomic features of multiple regions within and surrounding the primary tumor in breast cancer patients prior to receiving neoadjuvant chemotherapy (NAC) and the efficacy of NAC. By integrating clinical and pathological parameters, a predictive model was constructed to provide an accurate basis for personalized treatment and precise prognosis in breast cancer patients.
METHODS: This retrospective study included 321 breast cancer patients who underwent NAC treatment at the Second Hospital of Lanzhou University from January 2019 to December 2024. According to post-operative pathological results, the patients were divided into pathological complete response (PCR) and non-pathological complete response (non-PCR) groups. Regions of interest were outlined on 2-D ultrasound images using Itk-snap software. The intra-tumor (Intra) region and 5 mm (Peri-5 mm), 10 mm (Peri-10 mm) and 15 mm (Peri-15 mm) the peri-tumoralregions were demarcated, with radiomics features extracted from each region. Patients were randomly divided into a training set (n = 224) and a validation set (n = 97) in a 7:3 ratio. All features underwent Z-score normalization followed by dimensionality reduction using t-tests, Pearson correlation coefficients and least absolute shrinkage and selection operator. Radiomics models for Intra, Peri-5 mm, Peri-10 mm, Peri-15 mm and the combined intra-tumoral and peri-tumoral regions (Intra-tumoral, Peri-tumoral, IntraPeri) were constructed using a random forest machine-learning classifier. The predictive performance of the models was assessed by plotting receiver operating characteristic curves and calculating the area under the curve (AUC). Additionally, calibration curves and decision curve analysis were plotted to evaluate the model's goodness of fit and clinical net benefit RESULTS: A total of 214 radiomics features were extracted from the intra-tumoral and multi-region peri-tumoral areas. Using the least absolute shrinkage and selection operator regression model, eight intra-tumoral radiomics features, eight peri-10 mm radiomics features and nine IntraPeri-10 mm radiomics features were selected as being closely associated with PCR. The AUC of the intra-tumoral model was 0.860 and 0.823 in the training and validation sets, respectively. The AUCs of the peri-5 mm, Peri-10 mm and Peri-15 mm models were 0.836, 0.854 and 0.822 in the training set, and 0.793, 0.799 and 0.792 in the validation set. Among them, the AUC of the IntraPeri-10 mm model in the validation set was 0.842 (95% confidence interval [CI]: 0.764-0.921), which was superior to the AUC of the IntraPeri-5 mm model (0.831; 95% CI: 0.758-0.914) and the IntraPeri-15 mm model (0.838; 95% CI: 0.761-0.917). The combined model based on IntraPeri-10 mm and clinical pathological parameters (HER-2, Ki-67) achieved an AUC of 0.869 (95% CI: 0.800-0.937). The Delong test showed that the AUC of the combined model was significantly superior to that of the other models. The calibration curve indicated that the combined model had a good fit, and decision curve analysis demonstrated that the combined model provided a better clinical net benefit.
CONCLUSION: The peri-10 mm region is the optimal predictive area for the tumor's surrounding tissue after NAC in breast cancer. The IntraPeri-10 mm model, incorporating clinical pathological parameters, performs better at predicting the efficacy of NAC in breast cancer and can accurately assess treatment response, offering valuable guidance for subsequent treatment decisions.
引用
- Wei C, Jia Y, Gu Y, et al. Predictive Analysis of Neoadjuvant Chemotherapy Efficacy in Breast Cancer Using Multi-Region Ultrasound Imaging Features Combined With Pathological Parameters. Ultrasound Med Biol. 2025. PMID: 40976768. doi:10.1016/j.ultrasmedbio.2025.08.027
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。