DeepHealth highlights retrospective AI study modeling fewer prostate biopsies
Published in Artificial Intelligence, Journals, Scanners and Imaging, Tools.
October 8, 2026 — DeepHealth highlighted a multicenter prostate MRI study published in European Radiology on September 26, 2026. The retrospective analysis examined 787 men in Germany, the Netherlands and the United States; 380 had clinically significant prostate cancer.
A modeled strategy combining radiologists’ PI-RADS assessments with AI risk classification could have avoided 149 biopsies, or 18.9%, while maintaining 98.4% sensitivity for clinically significant disease. The strategy focused on uncertain PI-RADS 3 findings.
The distinction between patient and lesion performance matters: AI met the study’s non-inferiority criterion for patient-level sensitivity, but failed it for lesion-level sensitivity. Radiologists therefore retain an essential role in identifying and mapping individual lesions.
These figures describe simulated biopsy decisions using previously collected data, rather than a prospective demonstration that patients safely avoided procedures. The findings support further evaluation of AI as clinical decision support.
Sources: DeepHealth announcement, October 8, 2026; Guenzel and colleagues, European Radiology, September 26, 2026.