| 徐海敏,戴瑶,马雨竹,帅鸽,张妤.MR T1WI瘤体及瘤周影像组学联合临床特征预测乳腺癌新辅助化疗疗效[J].中国医学影像技术,2023,39(10):1520~1525 |
| MR T1WI瘤体及瘤周影像组学联合临床特征预测乳腺癌新辅助化疗疗效 |
| MR T1WI intratumoral and peritumoral radiomics combined with clinical features for predicting effect of neoadjuvant chemotherapy for breast cancer |
| 投稿时间：2023-05-22 修订日期：2023-09-17 |
| DOI：10.13929/j.issn.1003-3289.2023.10.016 |
| 中文关键词: 乳腺肿瘤 磁共振成像 影像组学 新辅助治疗 |
| 英文关键词:breast neoplasms magnetic resonance imaging radiomics neoadjuvant therapy |
| 基金项目: |
|
| 摘要点击次数: 1057 |
| 全文下载次数: 700 |
| 中文摘要: |
| 目的 观察MR T1WI瘤体和瘤周影像组学联合临床特征预测新辅助化疗(NAC)疗效的价值。方法 回顾性分析110例接受NAC的乳腺癌患者,其中43例NAC后病理完全缓解(pCR)、67例为非pCR(non-pCR);按7∶3比例将其分为训练集(n=76,30例pCR、46例non-pCR)和测试集(n=34,13例pCR、21例non-pCR)。以单因素及多因素logistic回归分析训练集临床及MRI表现,筛选NAC用于乳腺癌疗效的独立预测因子,并建立临床模型;于训练集NAC前MR T1WI所示瘤体及瘤周感兴趣体积(VOI)提取并筛选最佳影像组学特征,构建NAC治疗乳腺癌效果预测模型,包括模型瘤体、模型瘤周及模型瘤体+瘤周;联合瘤体及瘤周影像组学及临床相关独立预测因子建立联合模型。采用受试者工作特征(ROC)曲线评估模型诊断效能。结果 淋巴结转移(OR=0.17)、人表皮生长因子受体2(OR=4.52)及孕激素受体表达(OR=0.20)均为临床相关独立预测因子(P均＜0.05)。于瘤体及瘤周VOI各选出4个最佳影像组学特征并构建相应模型。联合模型在训练集的AUC为0.91,高于临床模型、模型瘤体、模型瘤周及模型瘤体+瘤周(AUC分别为0.85、0.72、0.72、0.74,P均＜0.05);其在测试集的AUC为0.88,高于模型瘤体(AUC=0.64,P＜0.05),与上述各模型的AUC(0.79、0.75、0.75)差异均无统计学意义(P均＞0.05)。结论 MR T1WI瘤周及瘤体影像组学联合临床特征可有效预测NAC治疗乳腺癌效果。 |
| 英文摘要: |
| Objective To investigate the value of MR T1WI intratumoral and peritumoral radiomics combined with clinical features for predicting effect of neoadjuvant chemotherapy (NAC) for breast cancer. Methods Data of 110 patients with breast cancer who underwent NAC were retrospectively analyzed, including 43 cases of pathological complete response (pCR) and 67 cases of non-pCR after NAC. The patients were divided into training set (n=76, 30 cases of pCR and 46 cases of non-pCR) or testing set (n=34, 13 cases of pCR and 21 cases of non-pCR) at the ratio of 7∶3. Univariate and multivariate logistic regression were used to analyze clinical and MRI findings of lesions in training set, and the independent predictors for effect of NAC were screened to establish clinical model. The best radiomics features based on MR T1WI intratumoral and peritumoral volume of interest (VOI) before NAC in training set were extracted and screened to construct predicting models, i.e. modeltumor, modelperitumor and modeltumor+peritumor. Then a combined model was established based on peritumoral and intratumoral radiomics combined with clinical features, and receiver operating characteristic (ROC) curves were drawn to evaluate the efficacy of the models. Results Lymph node metastasis (OR=0.17), human epidermal growth factor receptor-2 (OR=4.52) and progesterone receptor expression (OR=0.20) were all clinically relevant independent predictors (all P＜0.05). Based on each MR T1WI intratumoral and peritumoral VOI, 4 best radiomics features were screened to construct models, respectively. AUC of the combined model in the training set was 0.91, higher than that of clinical model, modeltumor, modelperitumor and modeltumor+peritumor (AUC=0.85, 0.72, 0.72, 0.74, all P＜0.05), in testing set was 0.88, higher than that of modeltumor (AUC=0.64, P＜0.05) but was not significant different from that of the other models (0.79, 0.75, 0.75, all P＞0.05). Conclusion MR T1WI intratumoral and peritumoral radiomics combined with clinical features could be used to effectively predict effect of NAC for breast cancer. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
