DWI-based Biologically Interpretable Radiomic Nomogram for Predicting 1-year Biochemical Recurrence after Radical...

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Biochemical recurrence (BCR) after prostatectomy for prostate cancer (PCa) is common.
• Early detection and management of BCR can improve PCa survival.
• Predicting BCR is crucial for timely intervention.

Purpose of the Study:

• To develop a nomogram integrating deep learning-based radiomic features and clinical parameters for predicting 1-year BCR after radical prostatectomy (RP).
• To explore the association between radiomic scores and the tumor microenvironment (TME).

Main Methods:

• Retrospective analysis of 349 patients undergoing RP with multiparametric MRI (mpMRI).
• Development of a 3D U-Net model for prostate cancer segmentation on diffusion-weighted imaging (DWI).
• Extraction of radiomic features and creation of predictive nomograms using multivariate Cox regression.
• Validation of nomogram performance, discrimination, calibration, and clinical usefulness.

Main Results:

• The clinical-radiomic nomogram achieved an AUC of 0.892 in the development cohort, outperforming radiomic and clinical models alone.
• The model demonstrated good performance and calibration in both development and validation cohorts (Hosmer-Lemeshow P > 0.05).
• Decision curve analysis confirmed superior clinical utility of the combined nomogram.

Conclusions:

• A DWI-based clinical-radiomic nomogram utilizing deep learning is feasible for predicting 1-year BCR after RP.
• Radiomic scores are linked to distinct tumor microenvironment patterns.
• This approach offers improved predictive accuracy and clinical usefulness for managing PCa recurrence.