Research Article
Framework to Segment and Evaluate Multiple Sclerosis Lesion in MRI Slices Using VGG-UNet
Table 3
Overall performance metrics computed for the test images of this study.
| CNN scheme | 2D plane | Jaccard | Dice | Accuracy | Precision | Sensitivity | Specificity |
| UNet | Sagittal | 82 ± 1.18 | 87 ± 2.61 | 98 ± 0.64 | 91 ± 2.92 | 88 ± 2.77 | 97 ± 1.08 | Axial | 81 ± 2.14 | 88 ± 3.06 | 98 ± 0.71 | 91 ± 3.52 | 88 ± 3.08 | 98 ± 0.51 | Coronal | 85 ± 2.29 | 88 ± 2.05 | 98 ± 0.32 | 91 ± 1.55 | 87 ± 4.16 | 97 ± 0.96 |
| SegNet | Sagittal | 82 ± 1.84 | 89 ± 1.16 | 97 ± 0.72 | 91 ± 2.37 | 87 ± 2.84 | 96 ± 1.38 | Axial | 83 ± 3.06 | 89 ± 1.55 | 96 ± 1.28 | 93 ± 2.16 | 88 ± 3.06 | 97 ± 0.52 | Coronal | 85 ± 1.16 | 89 ± 2.73 | 98 ± 0.51 | 91 ± 1.37 | 87 ± 3.33 | 97 ± 0.68 |
| VGG-UNet | Sagittal | 83 ± 1.22 | 91 ± 2.47 | 98 ± 0.38 | 91 ± 1.57 | 89 ± 3.03 | 98 ± 0.49 | Axial | 88 ± 1.06 | 92 ± 2.18 | 98 ± 0.18 | 90 ± 2.55 | 93 ± 1.68 | 98 ± 0.52 | Coronal | 89 ± 1.84 | 92 ± 3.17 | 98 ± 0.22 | 93 ± 2.07 | 94 ± 2.49 | 99 ± 0.06 |
| VGG-SegNet | Sagittal | 82 ± 2.05 | 90 ± 2.22 | 98 ± 0.27 | 90 ± 2.07 | 89 ± 2.55 | 98 ± 0.28 | Axial | 87 ± 2.53 | 91 ± 2.71 | 98 ± 0.31 | 90 ± 2.32 | 91 ± 2.16 | 98 ± 0.14 | Coronal | 87 ± 3.15 | 91 ± 1.45 | 98 ± 0.16 | 91 ± 2.17 | 92 ± 1.18 | 98 ± 0.23 |
| VGG-UNet (proposed) | Sagittal | 85 ± 3.05 | 92 ± 2.18 | 99 ± 0.04 | 92 ± 2.17 | 95 ± 0.48 | 99 ± 0.03 | Axial | 91 ± 2.21 | 94 ± 1.05 | 99 ± 0.10 | 94 ± 1.26 | 98 ± 0.12 | 99 ± 0.11 | Coronal | 90 ± 2.62 | 93 ± 1.11 | 99 ± 0.03 | 93 ± 1.94 | 97 ± 1.05 | 99 ± 0.08 |
|
|