Table 3.
Results of the classification analysis for cortical and lacunar stroke patients. The AUC values computed by averaging the results of the validation data (mean ± standard deviation) are shown for the two models (SVM with linear kernel and RF) and for all the MRI sequences and brain tissues/structures when using the texture features extracted from the 5 texture analysis methods. The presented values are obtained for the best tuning parameter in each case.
| AUC: Mean (SD) | GLRLM |
GLCM |
LBP |
WCF |
WSF |
|||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NAWM | SS | WMH | NAWM | SS | WMH | NAWM | SS | WMH | NAWM | SS | WMH | NAWM | SS | WMH | ||
| FLAIR | RF | <0.6 | <0.6 | <0.6 | <0.6 | <0.6 | <0.6 | <0.5 | <0.6 | <0.6 | <0.5 | 0.600 (0.118) | <0.5 | <0.5 | <0.5 | <0.5 |
| SVM | <0.6 | <0.6 | <0.5 | <0.5 | <0.6 | <0.5 | <0.6 | 0.604 (0.121) | <0.6 | <0.6 | <0.5 | <0.5 | <0.6 | <0.5 | <0.5 | |
| T2W | RF | <0.6 | <0.6 | <0.6 | <0.6 | 0.611 (0.121) | <0.5 | <0.5 | <0.5 | 0.616 (0.107) | <0.5 | <0.6 | <0.6 | <0.4 | 0.605 (0.117) | <0.4 |
| SVM | <0.5 | 0.622 (0.125) | <0.6 | 0.667 (0.117) | <0.5 | <0.5 | <0.6 | <0.5 | <0.6 | <0.6 | 0.604 (0.129) | <0.6 | 0.621 (0.140) | <0.6 | 0.661 (0.132) | |
| T1W | RF | <0.5 | <0.5 | <0.5 | <0.6 | <0.6 | <0.6 | <0.6 | <0.6 | <0.5 | <0.6 | <0.6 | <0.5 | 0.649 (0.114) | <0.6 | <0.5 |
| SVM | <0.5 | <0.5 | <0.6 | <0.6 | 0.637 (0.140) | <0.6 | <0.6 | <0.6 | 0.616 (0.092) | <0.6 | <0.5 | <0.5 | 0.618 (0.128) | <0.5 | <0.6 | |
AUC: area under the curve, RF: random forest classifier, SVM: support vector machine classifier, GLRM: grey–level run length matrix features, GLCM: grey-level co-occurrence matrix features, LBP: local binary patterns features, WCF: wavelet co-occurrence features, WSF: wavelet statistical features, NAWM: normal-appearing white matter, SS: subcortical structures, WMH: white matter hyperintensities.