Diagnosis of Clinically Significant Prostate Cancer in Multiparametric MRI with Pseudo-Localization of Suspected Lesion
In brief
AI model using pseudo-localization reaches 83% AUC for clinically significant prostate cancer
In a study of 1,494 mpMRI scans, the AUX-Net system combined with the PULSE algorithm achieved an AUC of 83.1%, with about 82% sensitivity and 76% specificity-about a two-point gain over the same model without pseudo-localization. The approach generates lesion locations automatically, letting unannotated exams train the AI, and higher localization quality linked to better diagnostic performance. Larger, prospective cohorts are needed before clinical adoption.
- Journal
- Physics in medicine and biology (Q1)
- Published
- 25 August 2026
- Study design
- Prospective / inception cohort
- Evidence level
- Level 2, Moderate (CEBM 2b)
- Authors
- Xijun Liu, Rongzong Liu, Xin Zhou, Yifei Yan, Haihao He, Quan Zhou, et al.
- PMID
- 42641644
- DOI
- 10.1088/1361-6560/ae9e7f
Why clinicians should know about it
- Picked for Radiology, Radiation Oncology, Nuclear Medicine, Medical Physics and Imaging (paper of the day, 28 August 2026): Pseudo‑localization improves csPCa diagnosis on mpMRI
Abstract
To develop a two-stage diagnostic framework using pseudo-localization for patient-level diagnosis of clinically significant prostate cancer (csPCa) on multiparametric MRI (mpMRI), enabling better utilization of examinations without manual lesion annotations.
Approach. We included 1494 mpMRI examinations from the PI-CAI dataset, partitioned at the patient level into 217 annotated csPCa, 202 unannotated csPCa, and 1075 non-csPCa cases. First, a prostate transformer U-Net (PTUnet) was trained on the 217 annotated cases via five-fold cross-validation to generate pseudo-locations for the 202 unannotated csPCa cases. These pseudo-locations, along with expert lesion annotations and non-csPCa labels, were used to train an anisotropic UX-Net (AUX-Net). The Probability-Ushered Lesion Selection (PULSE) procedure derived patient-level diagnostic probabilities. We evaluated PTUnet, AUX-Net, PULSE, and the pseudo-localization strategy against representative models and baselines, and analyzed the link between pseudo-localization quality and diagnostic performance.
Main results. PTUnet achieved a Dice similarity coefficient (DSC) of 64.96±4.22% and an average precision of 65.69±5.45%, the best overall pseudo-localization performance among evaluated models. All four diagnostic models showed higher AUCs with pseudo-locations. AUX-Net with PULSE achieved an AUC of 83.11%, sensitivity of 82.35%, specificity of 75.75%, and Youden's index of 58.10%, compared with the 80.96% AUC of AUX-Net without pseudo-locations. Pseudo-locations from nine different models all improved diagnostic AUC over the no-pseudo-location baseline, indicating the benefit is not model-specific. Furthermore, localization DSC was significantly and positively correlated with diagnostic AUC (Spearman's ρ = 0.800, p = 0.0096), suggesting higher-quality pseudo-localization leads to better diagnosis.
Significance. The framework allows csPCa examinations without manual annotations to contribute to diagnostic model training. Results show pseudo-localization improves patient-level diagnosis across different architectures, and its quality is positively associated with downstream performance. The framework shows potential for non-invasive AI-assisted csPCa diagnosis and biopsy triage, pending further validation on independent larger-scale datasets and prospective clinical cohorts.
Abstract as published, via PubMed.
For healthcare professionals. The summary is generated by AI from the published abstract, and the evidence level is assigned automatically from the study design on the Oxford CEBM hierarchy. Neither is medical advice. Read the full paper before changing practice.