AI and Cancer Research ── Prostate ── 2026-11-08
Prostate cancer: treatment, trials and AI
- Abstract
The objective was to evaluate the clinical context-dependent value of biparametric MRI (bpMRI)-derived imaging representations for prostate cancer (PCa) risk stratification, with emphasis on whether imaging features provide complementary information in patients within the prostate-specific antigen (PSA) gray zone. This dual-campus retrospective study included 536 patients with pathologically confirmed PCa who underwent pretreatment bpMRI. Patients were classified as low/intermediate and high-risk according to NCCN criteria. Patients from Campus 1 were randomly divided into training and internal test cohorts, and Campus 2 served as an independent test cohort. Radiomics, vision transformer, and contrastive language-image pretraining (CLIP) features were extracted from T2-weighted imaging, diffusion-weighted imaging, and apparent diffusion coefficient maps. Clinical and combined clinical-imaging models were developed. Performance was assessed using ROC analysis, calibration, decision curve analysis, and SHAP interpretation, with predefined evaluation in the PSA gray zone. In the independent test cohort, the combined model achieved the highest overall discrimination, with an AUC of 0.921, followed closely by the clinical baseline model (AUC, 0.907). Among imaging-only models, the radiomics model showed the highest overall AUC (0.844). SHAP identified PSA as the dominant predictor, indicating clinically driven overall performance. In the PSA gray zone, clinical and combined model performance declined (AUCs, 0.381 and 0.600), whereas CLIP achieved the highest AUC (0.785) and favorable net benefit. The value of bpMRI-derived imaging representations for PCa risk stratification is context-dependent. CLIP-derived representations may provide complementary discrimination when PSA-based information is less informative, supporting context-aware evaluation of imaging AI models.
Journal IF-equivalent: 3.9 (OpenAlex 2-year mean citedness, value as of 2026-10-13, retrieved 2026-10-15; not the official Clarivate IF)Reference: Zou F, Xiong H, Chen J, Qiao X, Ai G, Wu Y, et al. Clinical Context-Dependent Value of bpMRI-Derived Imaging Representations for Prostate Cancer Risk Stratification. J Digit Imaging Inform med. 2026 Nov 5 [Epub ahead of print]. doi:10.1007/s10278-026-02490-1.Checked: Abstract only - Abstract
This systematic review and meta-analysis evaluates the performance of magnetic resonance imaging (MRI)-based artificial intelligence (AI) models for predicting biochemical recurrence (BCR) in patients with prostate cancer treated with definitive external beam radiation therapy (EBRT), with or without concomitant hormone therapy. The review focuses on machine-learning, deep-learning, and radiomics-based prediction models that use MRI-derived information, either alone or in combination with clinical, pathological, or laboratory variables. The primary objective is to synthesize model performance using sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), while examining between-study heterogeneity and the robustness of pooled estimates. The review follows PRISMA 2020 and PRISMA-DTA reporting guidance, with data extraction informed by the CHARMS checklist for prediction-model studies.
Journal IF: unknown (could not be matched)Reference: [No authors listed]. Artificial Intelligence–Assisted Magnetic Resonance Imaging for Predicting Biochemical Recurrence After External Beam Radiation Therapy in Prostate Cancer: A Systematic Review and Meta-Analysis. Open Science Framework. [Epub ahead of print]. doi:10.17605/osf.io/kmh3s.Checked: Abstract only