Apparent diffusion coefficient radiomics for differentiating benign and malignant PI-RADS 3-5 prostate lesions: external validation.
Authors
Affiliations (4)
Affiliations (4)
- The Fifth Clinical Medical College of Henan University of Chinese Medicine (Zhengzhou People's Hospital), Zhengzhou, 450053, China. [email protected].
- Fugou Rehabilitation Hospital, Fugou, 461300, China.
- Fugou County People's Hospital, Fugou, 461300, China.
- The Second Hospital of Dalian Medical University, Dalian, 116000, China.
Abstract
To develop and externally validate a simplified apparent diffusion coefficient (ADC)-based MRI radiomics model for differentiating benign and malignant Prostate Imaging Reporting and Data System (PI-RADS) 3-5 lesions and assess its incremental value beyond clinical predictors. This retrospective multicenter study included 340 patients assigned to training (n = 182), internal test (n = 71), and external validation (n = 87) cohorts. Radiomics features were extracted from ADC maps and underwent reproducibility filtering and training-only feature selection. Four machine-learning classifiers were compared using nested cross-validation. Prostate-specific antigen (PSA), PSA density (PSAD), and PI-RADS were evaluated individually, and clinical and combined clinical-radiomics models were constructed. Logistic regression was selected as the final radiomics classifier. The radiomics model achieved area under the receiver operating characteristic curve (AUC) values of 0.712 and 0.679 in the internal test and external validation cohorts, respectively. Corresponding AUCs were 0.775 and 0.777 for the clinical model and 0.829 and 0.808 for the combined model. The combined model did not significantly outperform the clinical model (Holm-adjusted P = 0.123 and 0.392, respectively) and showed an external specificity of 0.538. ADC radiomics provided complementary information for differentiating benign and malignant PI-RADS 3-5 prostate lesions, but its incremental value beyond the clinical model was limited. Further prospective multicenter validation is warranted before clinical implementation.