Follicle detection and ovary detection results on the USOVA3D database testing set by applying different computer algorithms. Algorithms were evaluated and ranked by using USOVA3D database verification/validation protocol (see article: B. Potočnik et al.: “Public Database for Validation of Follicle Detection Algorithms on 3D Ultrasound Images of Ovaries”, Computer Methods and Programs in Biomedicine 196 (2020) 105621, doi:10.1016/j.cmpb.2020.105621 ).
OVARIAN FOLLICLE DETECTION
Effectiveness of various ovarian follicle detection methods evaluated on the USOVA3D database testing set. Final score statistics and the overall algorithm score are presented.
| Algorithm | median (\(\xi_{volume}\)) | min (\(\xi_{volume}\)) | max (\(\xi_{volume}\)) | \(\boldsymbol{\xi_{algorithm}}\) |
| EXT 1 + Deep Supervision | 80.9 | 63.2 | 93.5 | 80.5 |
| EXT 2 + Deep Supervision | 82.8 | 65.3 | 92.5 | 80.4 |
| 3D U-Net + Deep Supervision | 83.0 | 67.5 | 93.3 | 80.0 |
| 3D DWT-based method (baseline 1) | 79.3 | 59.7 | 90.6 | 78.2 |
| 3D U-Net | 76.0 | 59.6 | 90.4 | 74.8 |
| CNN-based method (baseline 2) | 75.1 | 43.8 | 91.5 | 72.5 |
OVARY DETECTION
Effectiveness of various ovary detection methods evaluated on the USOVA3D database testing set. Final score statistics and the overall algorithm score are presented.
| Algorithm | median (\(\xi_{volume}\)) | min (\(\xi_{volume}\)) | max (\(\xi_{volume}\)) | \(\boldsymbol{\xi_{algorithm}}\) |
| EXT 1 + Deep supervision | 79.9 | 51.6 | 91.5 | 77.7 |
| EXT 2 + Deep Supervision | 80.9 | 52.1 | 93.4 | 76.0 |
| 3D U-Net | 75.8 | 47.7 | 88.4 | 72.8 |
| CNN-based method (baseline 2) | 73.6 | 40.5 | 87.9 | 72.2 |
| 3D U-Net + Deep Supervision | 73.7 | 47.7 | 91.1 | 70.8 |
| 3D DWT-based (baseline 1) | 72.5 | 18.3 | 87.1 | 63.3 |