Penile parameters for circumcision
| Article DOI | https://doi.org/10.1016/j.jpedsurg.2025.162358 |
| Objective | To accurately identify normal penile parameters for circumcision eligibility from digital images of the penis using a mobile application. |
| AI Approach | CNN with transfer learning |
| Data Source(s) | Single institution image series (633 smartphone photos) |
| Model Input | 224 x 224 px color JPEG images of penis (dorsal, ventral, lateral) |
| Model Outcome | Eligibility for circumcision |
| Model Metrics | Dorsal: accuracy 89%, precision 89%, sensitivity 80%, specificity 94%, F1 0.88 Lateral: accuracy 91%, precision 89%, sensitivity 94%, specificity 87%, F1 0.91 Ventral: accuracy 93%, precision 96%, sensitivity 92%, specificity 93%, F1 0.92 |
| Model Usability | Deployed as the “Circapproved” feature inside the publicly available android app. Source code and data not publicly available. |
AI = Artificial intelligence
AIPEDURO 
Leave a comment