摘要
目的探讨基于双参数磁共振成像影像组学特征构建支持向量机(SVM)及随机森林(RF)两种机器学习模型预测前列腺癌风险分级的诊断作用。方法回顾性纳入经病理确诊为前列腺癌患者119例,其中中低危组57例,高危组62例,入组患者均在术前2个月内行MRI检查。分别提取基于T_(2)WI、ADC序列的影像组学特征。将入组患者按7∶3比例随机分为训练组和测试组。根据筛选后的影像组学特征分别建立基于T_(2)WI、ADC、T_(2)WI+ADC的SVM模型及RF模型,用测试组对模型进行验模型验证,检验每一种模型的准确率、特异性、敏感性并绘制受试者操作特征曲线(ROC)。采用曲线下面积(AUC)评估影像组学模型对前列腺癌风险分级的预测效能。结果基于T_(2)WI序列建立的SVM模型、RF模型的AUC分别为0.797、0.713;基于ADC序列建立的SVM模型、RF模型的AUC分别为0.826、0.667;T_(2)WI+ADC序列建立SVM模型、RF模型的AUC分别0.871、0.724。联合双参数的模型预测效能优于单参数模型。结论本研究构建的基于双参数磁共振的SVM及RF模型在一定程度上能预测前列腺癌风险分级,其中T_(2)WI+ADC的SVM模型分类效果更佳,有潜力应用于临床以指导前列腺癌患者的个体化治疗。
Objective To assess the diagnostic value of two machine learning models,support vector machine(SVM)and random forest(RF),based on the radiomic features from biparameter magnetic resonance imaging for predicting the risk classification of prostate cancer.Methods A total of 119 patients with histopathologically confirmed prostate cancer were retrospectively collected,including low-and medium-risk 57 patients and 62 high risk patients,all of whom were examined by magnetic resonance imaging(MRI)within 2 months before the operation.The radiomic features were extracted from T_(2)WI and ADC images.The enrolled patients were divided into a training group and a test group in a 7∶3 ratio.The SVM and RF models were establish based on the radiomic features.The models were validated by the test group to check the accuracy,specificity,and sensitivity of each model and to plot the subject operating characteristic curve(ROC).Finally,the area under the curve(AUC)was used to assess the predictive effectiveness of different models for prostate cancer risk classification.Results The AUCs of the SVM model and RF model based on the T_(2)WI sequence were 0.797 and 0.713.The AUCs of the SVM model and RF model based on the AUC sequence were 0.826 and 0.667.And the AUCs of the SVM model and RF model based on the T_(2)WI+ADC sequence were 0.871 and 0.724,respectively.The prediction performance of the biparameter model was better than that of the single-parameter model.Conclusion The biparameter MRI-based SVM and RF models can predict the risk classification of prostate cancer to some extent,the SVM model based on T_(2)WI+ADC has the best classification effect among them and the potential for clinical application to guide the individualized treatment of prostate cancer patients.
出处
《浙江临床医学》
2023年第5期656-659,共4页
Zhejiang Clinical Medical Journal
关键词
前列腺癌
影像组学
双参数磁共振
风险分级
Prostate cancer
Radiomics
Biparameter magnetic resonance imaging(bp-MRI)
Risk classification
