基于双模态MRI影像组学术前预测浸润性乳腺癌腋窝淋巴结转移

目的探讨T2WI联合DCE-MRI的影像组学特征术前预测浸润性乳腺癌腋窝淋巴结转移的价值。方法回顾性分析经手术病理证实的168例浸润性乳腺癌病人的临床病理资料及MRI图像资料。根据手术病理结果，将其分为淋巴结转移组（n=64）和无淋巴结转移组（n=104），并按8:2的比例将病人随机分为训练组（n=134）与验证组（n=34）。在T2WI和DCE两个序列手动勾画ROI进行图像分割和影像组学特征提取，利用Select K Best、LASSO回归及迭代筛选特征对高维组学特征进行降维，保留与腋窝淋巴结转移高度相关的特征。采用logistic回归建立T2WI、DCE和T2WI联合DCE三个影像组学预测模型，利用ROC曲线下面积（AUC）评估模型的效能，并以最优模型生成列线图。结果T2WI、DCE和T2WI联合DCE的影像组学预测模型在训练组的AUC分别为0.75、0.75和0.80；验证组的AUC分别为0.75、0.73和0.79。T2WI联合DCE模型的预测效能最佳。结论T2WI联合DCE影像组学预测模型在术前对浸润性乳腺癌腋窝淋巴结转移的预测具有一定的价值，能够无创、准确地预测腋窝淋巴结转移状态。

The preoperative prediction value based on dual-mode MRI image omics in axillary lymph node metastasis of invasive breast cancer

ObjectiveTo investigate the value of T2WI combined with DCE-MRI radiomics features in the preoperative prediction of axillary lymph node metastasis of invasive breast cancer.MethodsThe clinicopathological and MRI data of 168 patients with invasive breast cancer confirmed by surgical pathology were retrospectively analyzed. According to the pathological results, the patients were divided into the lymph node metastasis group(n=64) and non-lymph node metastasis group(n=104), and randomly divided into the training group(n=134) and verification group(n=34) in an 8:2 ratio. The ROI was manually delineated on T2WI and DCE sequences for image segmentation and image omics feature extraction. The Select K Best, LASSO regression and iterative screening features were used to reduce the dimensionality of high-dimensional omics features and retain the high associated features with axillary lymph node metastasis. The three imaging omics prediction models of T2WI, DCE and T2WI combined WITH DCE were established using the logistic regression. The area under the ROC curve(AUC) was used to evaluate the effectiveness of models, and the optimal model was used to generate a column chart.ResultsThe AUC of T2WI, DCE, and T2WI combined with DCE in the training group was 0.75, 0.75, 0.80, respectively. The AUC of T2WI, DCE, and T2WI combined with DCE in the validation group was 0.75, 0.73 and 0.79, respectively. The predictive performance in T2WI combined with DCE predictive model was the best.ConclusionsThe predictive model of T2WI combined with DCE has certain value in the preoperative prediction of axillary lymph node metastasis of invasive breast cancer. It can accurately and noninvasively predict the status of axillary lymph node metastasis.