Multicentre Evaluation of an AI-Assisted Urine Test for Clinically Significant Prostate Cancer in Men Undergoing Initial Biopsy.

Tags

Journal: Journal of extracellular vesicles

This study evaluates a non-invasive, urine-based test—Extracellular Vesicles Gene-based Prostate Score (EGPS)—to improve detection of clinically significant prostate cancer (csPCa) and reduce unnecessary biopsies in men with PSA 0–15 ng/mL undergoing initial biopsy.

Design and methods

  • Retrospective multicenter study: total n = 645 men.
  • Training and internal validation cohorts: 586 patients from three centers, split into 70% training and 30% internal validation.
  • External validation cohort: 59 patients from two additional centers.
  • Sample processing: urinary extracellular vesicles isolated using the EXODUS platform; gene expression quantified by RT-qPCR.
  • Model development: ten machine learning algorithms compared; the final EGPS model used three genes (AMACR, HOXB13, and PSGR).
  • Performance assessment: ROC curves, DeLong tests, and decision curve analysis.
  • AI integration: an AI diagnostic system based on DeepSeek was built to integrate EGPS for clinical use.
  • Sample collection: no digital rectal examination (DRE) was required.

Key results

  • Discrimination for csPCa (AUC):
  • Training cohort: 0.838.
  • Internal validation cohort: 0.825.
  • External validation cohort: 0.811.
  • Comparison with PSA: EGPS outperformed PSA in all cohorts.
  • Prespecified threshold of 0.22:
  • Sensitivity: exceeded 95% in all cohorts.
  • Missed csPCa: 3.81% in the training cohort and 0% in both validation cohorts.
  • Biopsy reduction while maintaining high sensitivity for csPCa:
  • Training cohort: avoided biopsies in 79 men (23.37%).
  • Internal validation cohort: avoided biopsies in 27 men (18.62%).
  • External validation cohort: avoided biopsies in 9 men (15.25%).

Clinical implications

  • DRE-free test: EGPS provides a urine-based, gene-expression test for men with PSA 0–15 ng/mL at initial biopsy evaluation.
  • Improved specificity: offers better specificity than PSA alone, with high sensitivity for csPCa.
  • Biopsy reduction: supports a meaningful reduction in unnecessary biopsies and associated risks.
  • AI-enabled decision-making: integration into a DeepSeek-based AI system supports early detection and more individualized biopsy decisions in a Chinese population, with potential for broader validation.

Leave a Reply