首页|期刊导航|分子影像学杂志|双参数MRI影像组学联合PI-RADS v2.1评分可有效提高前列腺特异性抗原灰区前列腺癌的诊断价值

双参数MRI影像组学联合PI-RADS v2.1评分可有效提高前列腺特异性抗原灰区前列腺癌的诊断价值OA

Biparametric MRI radiomics combined with PI-RADS v2.1 score can effectively improve the diagnosis value of prostate cancer in the PSA gray zone

中文摘要英文摘要

目的 探讨T2加权成像(T2WI)与弥散加权成像(DWI)序列的影像组学模型联合前列腺影像报告和数据系统版本2.1(PI-RADS v2.1)评分诊断前列腺特异性抗原(PSA)灰区(4~10 ng/mL)前列腺癌(PCa)的价值.方法 回顾性分析蚌埠医科大学第一附属医院2020年1月~2025年9月PSA在灰度区间119例患者的临床及影像资料,所有患者均经病理活检,其中PCa 35例,良性前列腺增生(BPH)84例,并于术前接受多参数磁共振(mpMRI)成像检查.将119例患者按8∶2比例随机分为训练集(95例,其中PCa 28例、BPH 67例)与测试集(24例,其中PCa 7例、BPH 17例),在T2WI及DWI图像上手动勾画感兴趣区(ROI),提取特征后行Min-max归一化,采用Pearson相关分析和Relief算法降维筛选特征,并用SMOTE均衡数据.训练集中以5折交叉验证分别构建基于支持向量机(SVM)和朴素贝叶斯的T2WI与DWI影像组学模型,融合双序列特征形成组学模型;经logistic回归筛选包括PI-RADS v2.1在内的临床独立预测因子建立临床模型,最终整合为联合模型,通过ROC曲线、列线图及决策曲线评估性能.结果 在测试集中,T2WI与DWI影像组学模型的曲线下面积(AUC)分别为0.790和0.807,准确度分别为0.792和0.667.双序列融合影像组学模型的诊断效能更优,其AUC为0.852(95%CI:0.775~0.910),但差异无统计学意义(P>0.05).将该融合模型与PI-RADS v2.1评分整合后,联合模型的AUC提升至0.870,诊断性能优于单一PI-RADS v2.1评分(0.781,P<0.01).结论 基于bpMRI的影像组学特征能有效量化肿瘤异质性,与PI-RADS v2.1评分的联合应用,能为PSA灰区前列腺癌的精准无创诊断提供更可靠的决策支持,有望减少不必要的穿刺活检.

Objective To investigate the value of radiomics models based on T2-weighted imaging(T2WI)and diffusion-weighted imaging(DWI)sequences combined with the Prostate Imaging Reporting and Data System version 2.1(PI-RADS v2.1)score in the diagnosis of prostate cancer(PCa)in the prostate-specific antigen(PSA)gray zone(4-10 ng/mL).Methods A retrospective analysis was conducted on the clinical and imaging data of 119 patients with PSA in the gray zone admitted to the First Affiliated Hospital of Bengbu Medical University from January 2020 to September 2025.All patients underwent pathological biopsy,including 35 cases of PCa and 84 cases of benign prostatic hyperplasia(BPH),and all received multiparametric magnetic resonance imaging(mpMRI)prior to surgery.The patients were randomly divided into a training set(95 cases,28 PCa and 67 BPH)and a test set(24 cases,7 PCa and 17 BPH)at an 8∶2 ratio.Regions of interest(ROIs)were manually delineated on T2WI and DWI images.After feature extraction,Min-max normalization was applied,and Pearson correlation along with the Relief algorithm were used for feature dimensionality reduction and selection.The synthetic minority over-sampling technique(SMOTE)was used to balance the data.In the training set,T2WI-and DWI-based radiomics models were built using support vector machine(SVM)and naive Bayes classifiers with 5-fold cross-validation.The features from the two sequences were fused to construct a combined radiomics model.Independent clinical predictors,including the PI-RADS v2.1 score,were screened by logistic regression to build a clinical model,which was ultimately integrated into a combined model.The diagnostic performance was assessed by ROC curve,nomogram,and decision curve analysis.Results In the test set,the AUC of the T2WI and DWI radiomics models were 0.790 and 0.807,with accuracies of 0.792 and 0.667,respectively.The biparametric MRI-based fusion radiomics model showed better diagnostic performance,with an AUC of 0.852(95%CI:0.775-0.910),but the difference was not statistically significant(P>0.05).When the fusion model was further integrated with the PI-RADS v2.1 score,the combined model achieved an AUC of 0.870,significantly outperforming the PI-RADS v2.1 score alone(0.781,P<0.01).Conclusion Radiomics features derived from bpMRI can effectively quantify tumor heterogeneity.Their integration with the PI-RADS v2.1 score may provide more reliable decision support for the accurate and non-invasive diagnosis of prostate cancer in the PSA gray zone,potentially reducing unnecessary biopsies.

周甜甜;周静;杨雨琼;李伟;马宜传;徐加利

蚌埠医科大学第一附属医院 放射科,安徽 蚌埠 233004||蚌埠医科大学研究生院,安徽 蚌埠 233030蚌埠医科大学第一附属医院 放射科,安徽 蚌埠 233004||蚌埠医科大学研究生院,安徽 蚌埠 233030蚌埠医科大学第一附属医院 放射科,安徽 蚌埠 233004||蚌埠医科大学研究生院,安徽 蚌埠 233030蚌埠医科大学第一附属医院 感染科,安徽 蚌埠 233004蚌埠医科大学第一附属医院 放射科,安徽 蚌埠 233004蚌埠医科大学第一附属医院 放射科,安徽 蚌埠 233004

前列腺癌前列腺特异抗原前列腺影像报告和数据系统版本2.1双参数磁共振成像影像组学

prostate cancerprostate-specific antigenProstate Imaging Reporting and Data System Version 2.1biparametric magnetic resonance imagingradiomics

《分子影像学杂志》 2026 (8)

990-997,8

安徽省教育厅重点项目(2025AHGXZK31539)

10.12122/j.issn.1674-4500.2026.08.02

评论