Department of Breast Cancer, Cancer Center, Guangdong Provincial People's Hospital and Guangdong Academy of Medical Sciences, Guangzhou, 510080, China.
The Second School of Clinical Medicine, Southern Medical University, Guangzhou, 510515, China.
Clin Transl Oncol. 2020 Jan;22(1):50-59. doi: 10.1007/s12094-019-02109-8. Epub 2019 Apr 11.
To evaluate the value of multiparametric magnetic resonance imaging (MRI) in pretreatment prediction of breast cancers insensitive to neoadjuvant chemotherapy (NAC).
A total of 125 breast cancer patients (63 in the primary cohort and 62 in the validation cohort) who underwent MRI before receiving NAC were enrolled. All patients received surgical resection, and Miller-Payne grading system was applied to assess the response to NAC. Grade 1-2 cases were classified as insensitive to NAC. We extracted 1941 features in the primary cohort. After feature selection, the optimal feature set was used to construct a radiomic signature using machine learning. We built a combined prediction model incorporating the radiomic signature and independent clinical risk factors selected by multivariable logistic regression. The performance of the combined model was assessed with the results of independent validation.
Four features were selected for the construction of the radiomic signature based on the primary cohort. Combining with independent clinical factors, the combined prediction model for identifying the Grade 1-2 group reached a better discrimination power than the radiomic signature, with an area under the receiver operating characteristic curve of 0.935 (95% confidence interval 0.848-1) in the validation cohort, and its clinical utility was confirmed by the decision curve analysis.
The combined model based on radiomics and clinical variables has potential in predicting drug-insensitive breast cancers.
评估多参数磁共振成像（MRI）在预测新辅助化疗（NAC）不敏感乳腺癌中的应用价值。
共纳入 125 例在接受 NAC 前接受 MRI 检查的乳腺癌患者（原发性队列 63 例，验证性队列 62 例）。所有患者均接受手术切除，并应用 Miller-Payne 分级系统评估对 NAC 的反应。将 1-2 级病例归类为 NAC 不敏感。在原发性队列中提取了 1941 个特征。经过特征选择后，使用机器学习构建了一个基于最优特征集的放射组学特征。我们建立了一个结合放射组学特征和多变量逻辑回归选择的独立临床危险因素的综合预测模型。通过独立验证的结果评估综合模型的性能。
基于原发性队列，选择了 4 个特征来构建放射组学特征。结合独立的临床因素，用于识别 1-2 级组的联合预测模型比放射组学特征具有更好的区分能力，在验证队列中的受试者工作特征曲线下面积为 0.935（95%置信区间 0.848-1），其临床实用性通过决策曲线分析得到了证实。
基于放射组学和临床变量的联合模型在预测药物不敏感乳腺癌方面具有潜力。
