Research Article
A Deep Learning Model for Stroke Patients’ Motor Function Prediction
Table 6
Floating-point operations per second and training CPU time of training of all methods.
| | Method | Average accuracy (%) | Average sensitivity (%) (percentage of patients with a dysfunction case who predicted as positive) | Average specificity (%) (percentage of patients without a dysfunction case who predicted as negative) |
| | EEG-DenseNet | 97.6% | 98.1% | 97.9% | | DenseNet | 92.4% | 93.2% | 92.8% | | Xception | 91.4% | 91.6% | 92.1% | | ResNet | 89.7% | 89.8% | 88.7% | | VGG16 | 87.4% | 88.2% | 88.6% |
|
|
