Fig. 1.
Fig. 2.
Fig. 3.
Fig. 4.
Fig. 5.
Fig. 6.
Fig. 7.
Fig. 8.
Performance of the proposed BTS-NEUNET method_
| Classes | Accuracy [%] | Precision [%] | Recall [%] | Specificity [%] | F1-Score [%] |
|---|---|---|---|---|---|
| Grey matter | 99.75 | 97.69 | 98.52 | 98.62 | 97.64 |
| White matter | 99.85 | 98.91 | 97.74 | 97.41 | 98.17 |
| Cerebrospinal fluid | 98.96 | 99.45 | 96.92 | 96.58 | 96.43 |
| Ischemic lesions | 99.69 | 99.17 | 98.28 | 97.34 | 97.49 |
| Healthy | 99.75 | 99.51 | 99.46 | 95.61 | 95.42 |
| Overall | 99.60 | 98.95 | 98.18 | 97.11 | 97.03 |
Comparison of a traditional network with the proposed DenseGoogLeNet_
| Technique | Accuracy [%] | Precision [%] | Recall [%] | Specificity [%] | F1-Score [%] |
|---|---|---|---|---|---|
| ShuffleNet [26] | 97.96 | 93.61 | 93.21 | 96.84 | 94.21 |
| ResNet [27] | 95.07 | 92.08 | 87.60 | 94.59 | 91.03 |
| Ghost Net [28] | 98.84 | 87.02 | 94.71 | 93.62 | 89.61 |
| MobileNet [29] | 95.66 | 94.42 | 92.35 | 97.85 | 95.87 |
| DenseGoogLeNet | 99.60 | 98.95 | 98.18 | 97.11 | 97.03 |
Accuracy comparison with the existing and the proposed method_
| Author | Technique | Accuracy [%] |
|---|---|---|
| Zhang, F., et la [16] | DDSeg | 97.68 |
| Kollem, S., et al [23] | Optimal SVM | 98.26 |
| Gudise, S., et al., [25] | CEFAFCM | 97.86 |
| Proposed method | BTS-NEUNET | 99.60 |