Deep learning based on multiparametric MRI for differentiating benign from malignant PI-RADS 3 prostate lesions: a dual-center study.
Authors
Affiliations (6)
Affiliations (6)
- Department of Radiology, School of Medicine, Taizhou First People's Hospital, Taizhou University, Taizhou, Zhejiang Province, China.
- Department of Radiology, Tongde Hospital of Zhejiang Province, Hangzhou, Zhejiang Province, China.
- Putuo People's Hospital, School of Medicine, Tongji University, Shanghai, China.
- Department of Pathology, School of Medicine, Taizhou First People's Hospital, Taizhou University, Taizhou, Zhejiang Province, China.
- Department of Radiology, School of Medicine, Taizhou First People's Hospital, Taizhou University, Taizhou, Zhejiang Province, China. [email protected].
- Department of Radiology, School of Medicine, Taizhou First People's Hospital, Taizhou University, No.218, Hengjie RoadHuangyan District, Taizhou City, 318020, Zhejiang Province, China. [email protected].
Abstract
To address the diagnostic challenge of indeterminate PI-RADS 3 prostate lesions, we developed a non-invasive deep learning radiomics model using multiparametric MRI. This dual-center retrospective study included 382 patients. Lesions on T2WI, DWI, and ADC were segmented and standardized. Deep learning features were extracted via ResNet101, and radiomics features were selected using minimum Redundancy Maximum Relevance (mRMR) and Least Absolute Shrinkage and Selection Operator (LASSO). We built and compared three models: clinical, deep-learning radiomics signature (DLRS), and a combined model. The combined model achieved the highest AUCs (0.951, 0.964, 0.925), significantly outperforming the clinical model (p < 0.05). The DLRS model also showed high performance (AUCs: 0.946, 0.932, 0.918). The nomogram was well-calibrated, and DCA indicated clinical utility. This mpMRI-based deep learning model accurately differentiates malignant from benign PI-RADS 3 lesions, offering a valuable non-invasive tool to aid clinical decisions and reduce unnecessary biopsies.