Table 2. Optimal patient response classifier parameter details, including texture features used.
| Classifier | Sensitivity
(%) |
Specificity
(%) |
Accuarcy
(%) |
AUC | F1-Score | Features | ROI Selection Method | Quantization | Pixel distance |
|---|---|---|---|---|---|---|---|---|---|
| FLD | 84 | 70 | 73 | 0.74 | 0.76 | MEA, VAR, STD | Margin (15 pixels) | 128 | 1 |
| SVM-Lin | 74 | 70 | 70 | 0.75 | 0.72 | MEA, STD | Core | 32 | 5 |
| SVM-RBF | 74 | 70 | 71 | 0.76 | 0.72 | MEA, STD, ENT, MAX | Core | 32 | 5 |
| kNN | 63 | 93 | 87 | 0.78 | 0.75 | HOM, ASM | Margin (10 pixels) | 128 | 1 |
Abbreviations: AUC: Area under curve; ROI: Region of interest; MEA: Mean; VAR: Variance; STD: Standard deviation; ENT: Entropy: MAX: Maximum; HOM: Homogeneity; ASM: Angular second moment.