ID | F1@50 | MCC | #TP | #FP | #FOG | %TF |
|---|
S1 | 87.5 | 95.7 | 9 / 9 | 0 / 24 | 10 / 9 | 37.1 / 33.9 |
S2 | 31.6 | 60.3 | 13 / 13 | 3 / 11 | 24 / 13 | 22.6 / 12.3 |
S3 | 60.0 | 59.8 | 3 / 5 | 2 / 23 | 3 / 5 | 5.34 / 6.89 |
S4 | 71.1 | 87.2 | 18 / 18 | 0 / 12 | 18 / 18 | 40.6 / 36.7 |
S5 | 85.7 | 96.9 | 3 / 3 | 1 / 23 | 3 / 3 | 35.2 / 36.1 |
S6 | 83.3 | 80.7 | 5 / 7 | 0 / 2 | 5 / 7 | 12.7 / 14.4 |
S7 | 100 | 98.5 | 1 / 1 | 0 / 30 | 1 / 1 | 19.5 / 20.1 |
| Â | 74.2 | 82.7 | 52 / 56 | 6 / 125 | 64 / 56 | 24.7 / 22.9 |
- Detailed overview of the FOG assessment performance of the proposed MS-GCN model for each subject. The fourth column depicts the number of true positive FOG detections (TP) with respect to the number of FOG episodes. The fifth column depicts the number of false-positive (FP) FOG detections with respect to the number of trials that did not contain FOG. The sixth and seventh columns depict the #FOG and %TF computed from the model annotated segmentations with respect to those computed from the expert annotated segmentations. All results were derived from the test set, i.e., subjects that the model had never seen
