MRI影像组学在鉴别前列腺癌与前列腺增生中的临床应用价值

目的:探讨基于T2WI与表观扩散系数(apparent diffusion coefficient,ADC)的影像组学模型对前列腺癌(prostate cancer,PCa)和前列腺增生(benign prostatic hyperplasia,BPH)的鉴别诊断应用价值。方法:回顾性分析经穿刺活检病理证实的72例PCa患者与64例BPH患者的临床资料及原始MRI图像。使用ITK-SNAP软件手动勾画整个肿瘤(不包括囊变、出血、坏死及钙化)生成三维体积(VOI)，采用FAE软件分别从T2WI与ADC图像提取高通量影像组学特征，应用Pearson相关性分析及循环特征消除(RFE)方法进行特征选择，使用支持向量机(SVM)作为分类器构建影像组学模型。采用分层随机抽样的方法将所有病例按7∶3分成训练集与验证集，通过绘制受试者工作特征(ROC)曲线及校准曲线分析评估各个影像组学模型的鉴别诊断效能与校准能力，并使用Delong检验评价模型间的差异。结果:经过特征选择，分别使用12、15、11个图像特征构建基于T2WI/ADC/T2WI+ADC的影像组学预测模型。训练集中T2WI/ADC/T2WI+ADC影像组学模型的ROC曲线下面积(AUC)分别为0.88、0.92、0.96，验证集中T2WI/ADC/T2WI+ADC影像组学模型的AUC分别为0.85、0.89、0.91。Delong检验显示T2WI+ADC联合影像组学模型鉴别诊断效能显著高于单个序列模型，校准曲线显示联合影像组学模型具有较好的预测能力。结论:基于T2WI+ADC的联合影像组学模型有助于术前鉴别PCa和BPH。

Clinical application value of MRI radiomics in differentiating prostate cancer from prostatic hyperplasia

Objective:To investigate the application value of radiomics model based on T2WI and apparent diffusion coefficient(ADC) in diagnosis of prostate cancer(PCa) and benign prostatic hyperplasia(BPH).Methods:The clinical data and original MRI images of 72 patients with PCa and 64 patients with BPH confirmed by needle biopsy pathology were retrospectively analyzed.The region of interest(ROI) of the tumor(excluding cystic change,hemorrhage,necrosis and calcification) was manually delineated using ITK-SNAP software to volume of interest(VOI).High-throughput radiomics features were extracted from T2WI and ADC images using FAE software.Pearson correlation method and recursive feature elimination(RFE) method were used for feature screening.Support vector machine(SVM) was used as a classifier to construct the radiomics model.All cases were divided into training set and verification set according to 7∶3 by stratified random sampling.The receiver operating characteristic(ROC) curve and calibration curve analysis were used to analyze and evaluate the diagnostic efficacy and calibration ability of each radiomics model,and the differences between the models were evaluated by Delong's test.Results:After feature selection,12,15 and 11 image features were used to construct radiomics models based on T2WI,ADC and their combination.The area under the ROC curve(AUC) of each prediction model in the training set was 0.88,0.92 and 0.96, respectively,and the AUC of each prediction model in the validation set was 0.85,0.89 and 0.91,respectively.Delong test showed that the diagnostic performance of T2WI+ADC combined radiomics model was significantly higher than that of single sequence model,and the calibration curve showed that the combined radiomics model had better predictive ability.Conclusion:The T2WI+ADC combined radiomics model can distinguish PCa from BPH more comprehensively,objectively and accurately.