Development and Validation of a Multimodal AI-Based Model for Predicting Post-Prostatectomy Treatment Outcomes from...

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Imaging

Background:

• Prostate cancer (PCa) is a leading cause of cancer death in men.
• Current risk prediction methods for PCa lack accuracy and reproducibility.
• Predicting disease severity and recurrence after treatment remains challenging.

Purpose of the Study:

• To develop and validate an automated multimodal artificial intelligence (AI) algorithm for predicting biochemical recurrence (BCR) after radical prostatectomy (RP) in PCa patients.
• To compare the performance of the AI model against clinical standards.
• To assess the AI model's ability to differentiate BCR-free survival outcomes, particularly in intermediate-risk groups.

Main Methods:

• An automated multimodal AI algorithm was developed using biparametric MRI (bpMRI) and clinical covariates.
• Data from two centers were used, including development (n=240), test (n=71), and external validation (n=168) cohorts.
• Radiomics features were extracted from bpMRI, and models were compared based on accuracy, sensitivity, specificity, and AUC.

Main Results:

• The multimodal AI model (M3) achieved the highest AUC across test sets (combined: 0.71; center 1: 0.70; center 2: 0.75).
• The multimodal model was the only one to significantly differentiate BCR-free survival outcomes in intermediate-risk groups across both centers (p < 0.05).
• The AI model's performance approached clinical gold standards.

Conclusions:

• An automated multimodal AI algorithm integrating radiomics and clinical data can effectively predict BCR after RP.
• This AI approach shows promise for enhancing imaging-based prognostication in prostate cancer.
• Further validation is warranted to fully integrate this tool into clinical practice.