基于增强MRI影像组学在乳腺癌腋窝淋巴结转移预测模型的研究

A radiomic model based on dynamic contrast enhanced MRI(DCE-MRI)for predicting axillary lymph node(ALN)metastasis in patients with breast cancer

目的探讨基于增强磁共振扫描影像组学特征构建乳腺癌腋窝淋巴结(ALN)转移预测模型的价值。方法研究纳入我院收治的100例经病理检查证实的乳腺癌患者(ALN转移55例,ALN未转移45例)。患者均行增强磁共振扫描并分别构建传统影像预测模型和影像组学预测模型,比较两种预测模型的预测乳腺ALN转移的价值。结果传统影像特征经Logistics多因素分析发现淋巴结门消失、淋巴结短径增加、DWI信号增强和ADC值降低是预测乳腺癌ALN转移的独立因子(P<0.05)。与传统影像特征相比,影像组学预测模型通过降维后得到长行程优势、全角度集群突出方差、均匀度、全角度相关性、表容比5个具有特征性参数,在预测ALN转移与未转移患者中存在差异有统计学意义(P<0.05);影像组学在ALN转移预测模型的AUC值、敏感度、特异性、准确性分别为0.944、0.900、0.883、0.889,明显高于传统影像ALN预测模型。结论基于增强磁共振扫描的影像组学特征和传统影像特征构建的乳腺癌ALN转移预测模型均具有一定的预测效能,且影像组学预测模型的诊断价值明显高于传统影像预测模型。

Objective To investigate the prognostic value of radiomics model based on dynamic contrast enhanced MR imaging(DCE-MRI)in predicting axillary lymph nodes(ALN)metastasis of breast cancer.Methods One hundred patients were diagnosed in breast cancer by pathology including 55 case of axillary lymph node positive and 45 of negative.All case were underwent conventional DCE-MRI before surgery.Radiomic features and hemodynamics characteristics were derived from DCE-MRI data and built two different model based on radiomic features and conventional DCE-MRI,compared the value of the two model on predicating of the metastasis of the axillary lymph nodes.Results the breast cancer ALN metastasis prediction model based on enhanced MRI features and traditional image features has certain predictive power,Moreover the diagnostic value of the imaging group prediction model is higher than that multivariate analysis revealed that the loss of lymph node,the increase of short diameter of lymph nodes,the enhancement of DWI signal and the decrease of ADC value were the independent factors predicting the ALN metastasis of breast cancer(P<0.05)by Logistics.Compared with the traditional image features,after dimensionality reduction,the imageomics prediction model obtained five characteristic parameters:long stroke advantage,full angle cluster prominent variance,evenness,full angle correlation and surface volume ratio,which were significantly different between patients with AlN metastasis and patients without AlN metastasis(P<0.05).The AUC value,sensitivity,specificity and accuracy of imageomics in AlN metastasis prediction model were 0.944,0.900,0.883 and 0.889 respectively,which were significantly higher than those of traditional image AlN prediction model.Conclusion the prediction model of ALN metastasis of breast cancer based on the imaging features of contrast-enhanced MRI and the conventional imaging features has certain prediction efficiency,Moreover the diagnostic value of the imaging features prediction model is higher than that of the conventional image prediction model.