Paige Prostate Detect

Histopathology prostate

AI-based detection of prostate cancer foci in core needle biopsies; first FDA-authorized AI software for prostate cancer pathology.

Regulatory

Regulatory status	De Novo authorized
FDA submission pathway	De Novo
FDA decision date	2021-09-21

Validation summary

Study design	Retrospective
Cohort size	527 patients; 527 samples — 527 whole slide images (171 cancer + 356 benign), one slide per unique patient; 16 pathologist readers
Number of validation sites	157
Site geography	Multi-center (international)
Comparator	Pathologist consensus
Primary endpoint	Sensitivity / specificity
Primary result	Combined (16 pathologists, assisted vs unassisted): sensitivity 96.8% vs 89.5% (improvement 7.3%, 95% CI 3.9–11.4%, statistically significant); specificity 89.5% vs 88.4% (difference 1.1%, 95% CI -0.7–3.4%, not statistically significant). Stand-alone algorithm localization & accuracy study (728 WSIs): sensitivity 94.5% (95% CI 91.4–96.6%), specificity 94.0% (95% CI 91.3–95.9%).
External validation	Yes — FDA pivotal reader study used 527 WSIs sourced from 1 internal US site (44.15%) plus 156 different external sites worldwide (55.85%); none of the WSIs overlapped with the algorithm development data. Algorithm localization/accuracy study (728 WSIs) included slides from 217 different external sites worldwide. · Sensitivity stratified by source — internal site 94.1% (88.8–97.0%), external sites 94.9% (90.5–97.3%). Specificity — internal site 96.7% (93.0–98.5%), external sites 91.9% (87.7–94.7%).
Peer-reviewed	Yes
FDA decision summary	View on accessdata.fda.gov

Key publications

Clinical Validation of Artificial Intelligence-Augmented Pathology Diagnosis Demonstrates Significant Gains in Diagnostic Accuracy in Prostate Cancer Detection pivotal
Archives of Pathology & Laboratory Medicine · 2023
Novel artificial intelligence system increases the detection of prostate cancer in whole slide images of core needle biopsies
Modern Pathology · 2020

Limitations noted

FDA Decision Summary notes: (i) the clinical study was 'on a per-biopsy basis, not on a per-patient basis. … the expected benefit of the use of the Paige device on the final diagnosis in practice would likely be substantially lower than 7.3% when evaluated on a per-patient basis'; (ii) the dataset 'was enriched with 50% challenging cancer slides, which were defined as slides with minimal tumor burden (≤0.5mm)' — performance may differ on routine-difficulty cases; (iii) initial interpretation only — 'special studies were not permitted'; (iv) all slides scanned with a single Philips Ultra Fast Scanner — performance with other scanners not evaluated; (v) the device is 'an adjunctive computer-assisted methodology and its output should not be used as the primary diagnosis.'

Deployment

Available

Yes

Sources

https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200080.pdf · fda_db · 2026-04-20

How to cite this entry

Elbakri A. OncologyAI Registry: Paige Prostate Detect. Available at: https://oncologyairegistry.org/tools/paige-prostate-detect.html. Accessed [date].