知识与财富的链接
目的 基于临床、MRI影像组学及深度学习(DL)构建联合模型,评估其预测初发前列腺癌(PCa)骨转移(BM)的价值。方法 回顾性分析286例经病理证实的初发PCa,根据患者来源将其分为训练集(53例BM、121例无BM)和测试集(29例BM、83例无BM)。采用单因素分析及多因素logistic回归分析筛选初发PCa BM的临床独立风险因素,构建临床模型;基于MR T2WI和弥散加权成像(DWI)提取并筛选最佳影像组学特征,构建影像组学标签评分(Rad-score),以最佳DL特征建立DL标签评分(DL-score),进而构建影像组学-DL模型;最后基于临床独立风险因素、Rad-score及DL-score构建联合模型。绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),评估各模型预测初发PCa BM的效能,以决策曲线分析(DCA)对比其临床获益。结果 血清前列腺特异抗原(PSA)(OR=1.003,P<0.01)及国际泌尿病理学会(ISUP)评分(OR=3.023,P=0.01)为初发PCa BM的临床独立风险因素;以之构建的临床模型预测训练集及测试集初发PCa BM的...
Objective To construct the combined model based on clinic, MRI radiomics and deep learning (DL), and to explore its value for predicting bone metastasis (BM) of initial prostate cancer (PCa). Methods Totally 286 patients with pathologically confirmed initial PCa were enrolled and divided into training set (53 cases of BM, 121 cases of non-BM) and test set (29 cases of BM, 83 cases of non-BM) according to research centers. Clinical independent risk factors for BM of initial PCa were screened using univariate analysis and multivariate logistic regression analysis, and then a clinical model was constructed. The best radiomics features were extracted and screened based on MR T2WI and diffusion weighted imaging (DWI) to construct the radiomics label score (Rad-score), and the best DL features were extracted and screened to construct DL label score (DL-score), then the radiomics-DL model was constructed. Finally the combined model was constructed based on clinical independent risk factors, Rad-score and DL-score. Receiver operating characteristic (ROC) curves were drawn, and the areas under the curve (AUC) were calculated to evaluate the efficacy of each model for predicting BM of initial PCa. Decision curve analysis (DCA) was applied to compare the clinical benefits of different models. Results Serum PSA (OR=1.003, P<0.01) and ISUP score (OR=3.023, P=0.01) were clinically independent risk factors for BM in initial PCa patients. AUC of clinical model constructed based on the above factors for predicting BM in initial PCa patients in training set and test set was 0.79 and 0.81, respectively, of radiomics-DL model and combined model in training set was 0.90 and 0.93, while in test set was 0.92 and 0.95, respectively. AUC of combined model for predicting BM of initial PCa in both training set (Z=3.12, P<0.01; Z=1.76, P=0.04) and test set (Z=2.89, P<0.01; Z=2.23、P=0.03) were higher than that of clinical model and radiomics-DL model. DCA showed that taken 0.10-0.78 as the threshold, clinical benefit of the combined model was higher than that of clinical model and radiomics-DL model. Conclusion Clinic, MRI radiomics and DL combined model could be used to effectively predict BM of initial PCa.