An evaluation of artificial intelligence assisted prostate biopsy reporting in the Articulate Pro study.
Browning L., Colling RT., Oxley J., Birks J., Alham NK., Malacrino S., Zhang C., Ghosh A., Cooper R., Dolton M., Horton M., Retamero JA., Protheroe A., Ahmed LB., Banu N., Chatzitolios A., Danial I., Al-Hyassat S., Sinha B., Borkar P., Gill P., Bryant RJ., Aning J., White K., Scheffer R., Stanislaus E., Crofts J., Snead DJ., Rajpoot N., Hutton K., Hewitt D., Dunstan C., Procter R., Verrill C.
Prospective evidence on clinical utility of AI in histopathology is limited. We conducted a prospective study across three National Health Service specialist centres in England to evaluate a commercially available AI system for assistance in prostate biopsy reporting. Of 1613 cases, 1049 were reported with AI-assistance. Endpoints evaluated diagnostic impact, clinical impact and workflow. Staged AI assistance (second-read) prompted case review and changed the initial diagnosis or Grade Group of 21/386(5.4%) patients, 5 of these (1.3%) potentially affecting clinical management. AI-assisted workflows showed significantly reduced mean turnaround time with concurrent-read compared to unassisted-read by 30.1 h (p < 0.0001) at one site with significant reductions in cases requiring immunohistochemistry in all sites (Odds Ratios 0.50,0.43,0.33, p < 0.0001, p = 0.01, p = 0.001). This first prospective, multi-centric evaluation demonstrates AI can enhance diagnostic accuracy, shorten turnaround times and reduce unnecessary testing. Scaled across the NHS, such improvements could improve patient care, deliver faster diagnoses and optimise laboratory efficiency, supporting adoption.