MRI影像组学预测非特殊型浸润性乳腺癌分子分型的价值

Value of MRI radiomics in predicting molecular subtypes of invasive breast carcinoma of no special type

目的 探讨基于MRI影像组学预测非特殊型浸润性乳腺癌Luminal型和非Luminal型的临床价值。材料与方法 回顾性分析本院2021年4月至2022年12月经病理证实为非特殊型浸润性乳腺癌的患者149例,均在治疗前两周进行了MRI平扫和增强扫描。收集全部入组患者的临床及病理资料,根据雌激素受体(estrogen receptor,ER)和孕激素受体(progesterone receptor,PR)的表达情况将患者分为Luminal型(n=90)和非Luminal型(n=59)。以7∶3的比例随机将其分为训练组(n=104)和测试组(n=45)。将提取的数据进行降维并筛选影像组学最优特征,基于随机森林法建立三个预测模型,分别是扩散加权成像(diffusion weighted imaging,DWI)序列模型、动态对比增强(dynamic contrast-enhanced,DCE)-MRI序列模型以及DWI和DCE序列联合模型。采用受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under the curve,AUC)评价模型的预测性能。不同模型的预测效能采用DeLong检验进行比较。结果 Luminal型和非Luminal型组间、训练组和测试组组间患者的临床病理特征(年龄、ER状态、PR状态、绝经状态、淋巴结转移情况)差异均无统计学意义(P>0.05)。DWI模型、DCE模型和联合模型在训练组中的AUC分别为0.859、0.839、0.903,在测试组中的AUC分别为0.722、0.798、0.821。DeLong检验显示训练组的DCE模型和联合模型预测效能差异有统计学意义(P=0.03),除此之外三个模型两两比较预测效能差异均无统计学意义(P>0.05)。结论 基于MRI影像组学构建的模型可以较好地预测非特殊型浸润性乳腺癌Luminal型和非Luminal型,并能为非特殊型浸润性乳腺癌临床治疗方案的决策提供帮助。

Objective:To investigate the clinical value of MRI radiomics in predicting Luminal and non-Luminal subtypes of invasive breast carcinoma of no special type.Materials and Methods:A total of 149 cases with pathologically confirmed invasive breast carcinoma of no special type from April 2021 to December 2022 were retrospectively collected according to the inclusion and exclusion criteria,and all of them underwent non-contrast MRI scanning and contrast-enhanced scanning before treatment.Clinical and pathological data of all enrolled patients were collected.Patients were classified into Luminal subtype(n=90)and non-luminal subtype(n=59)based on the expression of estrogen receptor(ER)and progesterone receptor(PR).They were divided into training group(n=104)and test group(n=45)in a ratio of 7∶3,randomly.Screened the optimal features of radiomics from the extracted data,and three prediction models were established based on the random forest method,namely DWI model,DCE model,and DWI combining DCE model.The prediction performance of the models was evaluated by receiver operating characteristic(ROC)curve and the area under the curve(AUC).The DeLong test was used to compare the prediction performance of different models.Results:There were no significant differences in the clinicopathological features(age,ER status,PR status,menopausal status,lymph node metastasis)between Luminal and non-Luminal groups,training group and test group(P>0.05).The AUCs of DWI model,DCE model and joint model were 0.859,0.839 and 0.903 in the training group.In the test group,the AUCs were 0.722,0.798 and 0.821 respectively.The DeLong test showed that the prediction efficacy of the DCE model and the joint model in the training group were statistically significant(P=0.03),but there was no statistically significant difference in the prediction efficacy among the three models(P>0.05).Conclusions:The model constructed based on MRI radiomics can better predict the luminal and non-luminal subtypes of invasive breast carcinoma of no special type,and can help the decision-making of clinical treatment options for invasive breast carcinoma of no special type.