Avenda Health AI Prostate Cancer Planning Software
AI-based decision-support adjunct to MRI + biopsy review for the prostate oncological workflow. Lesion characterization functions intended for use on patients with pathology-confirmed Gleason Grade Group (GGG) ≥ 2 lesion and biopsy coordinates uploaded — used to evaluate the extent of known disease. The device generates a 3D voxel-level cancer probability map and an Encapsulation Confidence Score for clinician-selected boundaries. Supports treatment-planning (biopsy, focal soft-tissue ablation). Also functions as a medical image viewer (3D-visualization, comparison) for prostate MRI. Marketed commercially as Unfold AI.
Regulatory
| Regulatory status | 510(k) cleared |
| FDA submission pathway | 510(k) |
| FDA decision date | 2022-11-22 |
| FDA 510(k) number | K221624 |
Validation summary
| Study design | Retrospective |
| Cohort size | 50 patients; 50 samples — FDA MRMC reader study used a whole-mount prostatectomy database of GGG 2-3 patients (representative of intended-use population). Each whole-mount sample included 3D surfaces and pathology labels registered to preoperative T2-weighted MRI. 10 practicing urologists or radiologists from different institutions (range 2–23 years of experience). Each reader reviewed each case. Standalone segmentation test set: 137 patients (prostate segmentation); 50 patients (whole-mount pathology lesion-contouring accuracy). |
| Number of validation sites | 5 |
| Site geography | Multi-center (US) |
| Comparator | Pathologist consensus |
| Primary endpoint | Sensitivity / specificity |
| Primary result | Reader study (lesion contours vs whole-mount pathology ground truth): Sensitivity mean 97.4% (with device) vs 38.2% (SOC), p<0.0001; Specificity mean 72.1% (with device) vs 53.4% (hemi-gland contours), p<0.0001; Balanced accuracy 84.7% (device) vs 67.2% (SOC) vs 75.9% (hemi-gland), p<0.0001; Complete csPCa encapsulation rate 72.8% (device) vs 1.6% (SOC), p<0.0001; clinical-quality rating: 99% device vs 60% hemi-gland (p<0.0001). |
| External validation | Yes — Reader-study cases were independent of training set per FDA summary. The peer-reviewed J Urol 2024 paper (Priester et al.) used a 50-case prostatectomy whole-mount cohort retrospectively eligible for focal therapy; 7 urologists + 3 radiologists from 5 institutions, experience 2–23 years. · Concordant with FDA summary results: AI assistance improved encapsulation accuracy ~45× vs SOC. |
| Peer-reviewed | Yes |
| FDA decision summary | View on accessdata.fda.gov |
Key publications
- Artificial Intelligence Improves the Ability of Physicians to Identify Prostate Cancer Extent pivotalJournal of Urology · 2024
- Prediction and Mapping of Intraprostatic Tumor Extent with Artificial IntelligenceEuropean Urology Open Science · 2023
Limitations noted
[Source-detail preserved: Study design: Multi-center MRMC (multi-reader, multi-case); Primary endpoint (original): sensitivity_specificity_lesion_contouring] (i) Indication is restricted to patients with pathology-confirmed GGG ≥ 2 lesion AND biopsy-coordinate upload — NOT a screening or unaided detection tool; (ii) test cohort 50 patients is small; (iii) predicate is QuantX (DEN170022) but Avenda is the first FDA device specifically for prostate cancer extent mapping; (iv) ground truth is whole-mount prostatectomy — generalizability to patients NOT going to prostatectomy not established; (v) reader pool included urologists in addition to radiologists, broader than typical radiology MRMC.
Deployment
| Available | Yes |
Sources
- https://www.accessdata.fda.gov/cdrh_docs/pdf22/K221624.pdf · fda_db · 2026-05-12
- https://pubmed.ncbi.nlm.nih.gov/38860576/ · publication · 2026-05-12
Elbakri A. OncologyAI Registry: Avenda Health AI Prostate Cancer Planning Software. Available at: https://oncologyairegistry.org/tools/avenda-prostate-cancer-planning.html. Accessed [date].