基于mp-MRI影像组学预测前列腺癌Gleason分级的价值探讨

目的探讨多参数磁共振成像（mp-MRI）影像组学预测前列腺癌Gleason分级的价值。方法回顾性分析266例前列腺癌病人，根据病理结果分为Gleason评分高危组（Gleason≥4+3分）、Gleason评分低危组（Gleason≤3+4分），在T2WI横断面、ADC图（b值0、1 500 s/mm2）上手动勾画病灶后进行影像组学特征的提取及量化，将所选病例数据特征随机分为训练组与测试组（测试集比例为0.3），构建支持向量机分类模型，得到训练组与测试组的ROC曲线及曲线下面积（AUC）。结果Gleason评分高危组118例，Gleason评分低危组148例，其中训练组186例（高危组83例、低危组103例），测试组80例（高危组35例、低危组45例），T2WI、ADC图影像组学支持向量机模型训练组的AUC为0.753，测试组AUC为0.741，准确率为62.5%（95%CI:0.572~0.893）；T2WI、ADC图影像组学联合PSA值支持向量机模型训练组的AUC为0.768，测试组AUC为0.752，准确率为72.5%（95%CI:0.613~0.917）；经Delong验证，两者差异无统计学意义（P>0.05）。结论mp-MRI影像组学预测前列腺癌Gleason分级有较高的参考价值。

Value of mp-MRI radiomics in predicting Gleason grading of prostate cancer

ObjectiveTo explore the value of mp-MRI radiomics in predicting the Gleason grading of prostate cancer.MethodsThe clinical data of 266 prostate cancer patients were retrospectively analyzed, and the patients were divided into the Gleason score high-risk group(Gleason ≥ 4+3 points) and low-risk group(Gleason ≤ 3+4 points) according to the pathological results.After the lesions were manually delineated on the T2WI cross-section and ADC map(b value 0, 1 500 s/mm2), the image features were extracted and quantified.The data characteristics of cases were randomly divided into the training group and test group(test set ratio for 0.3), and the support vector machine classification model was constructed to obtain the ROC curve and area under the curve(AUC) of training group and test group.ResultsThere were 118 cases in the Gleason score high-risk group and 148 cases in the Gleason score low-risk group, and there were 186 cases in the training group(including 83 cases in the high-risk group and 103 cases in the low-risk group) and 80 cases in the test group(including 35 cases in the high-risk group and 45 cases in the low-risk group).The AUC value of the training group and test group in the T2WI and ADC radiomics support vector machine model were 0.753 and 0.741, respectively, and the accuracy rate of test group was 62.5%(95%CI: 0.572-0.893).The AUC value of the training group and test group in the T2WI, ADC radiomics combined with PSA value support vector machine model were 0.768 and 0.752, respectively, and the accuracy rate of test group was 72.5%(95%CI: 0.613-0.917).The results of Delong verification showed that the difference between the two was not statistically significant(P>0.05).ConclusionsThe mp-MRI radiomics has a high reference value in predicting the Gleason grading of prostate cancer.