Drainage in ureteropelvic junction obstruction, Hydronephrosis.
| Article DOI | https://doi.org/10.1016/j.jpurol.2022.12.017 |
| Objective | To analyze MAG3 renal scans using machine learning to predict renal complications in prenatal hydronephrosis |
| AI Approach | 2 model approaches: 1. CNN 2. Random forrest (RF) |
| Data Source(s) | Single institutional series (152 patients) |
| Model Input | CNN: Radiotracer concentration on MAG3 study images, over time RF: Age, sex, race, laterality, differential function measurements, T1/2 |
| Model Outcome | Renal complications (defined as decline of greater than 5% in renal function of the affected kidney, increased parenchymal thinning on US, worsening hydronephrosis, new onset flank pain, or occurrence of pyelonephritis) |
| Model Metrics | CNN: AUC: 0.78 Accuracy: 73% RF: AUC: 0.67 Accuracy: 0.62 |
| Model Usability | NA |
AI = Artificial intelligence, SFU = Society for Fetal Urology, CNN = Convolutional Neural Network
AIPEDURO 
Leave a comment