TABLE VI.
Summary of average number of image features used in 5 different SVM models and classification performance (AUC) based on both region and case-based lesion classification. p value compares results of each model to the last one (RPA) as the optimal one.
| Feature sub-groups |
Number of features |
AUC |
p value |
|---|---|---|---|
| Original features, region based | 181 | 0.72 | 0.004 |
| Original features, case based | 181 | 0.74 | 0.005 |
| NMF, region based | 100 | 0.73 | 0.005 |
| NMF, case based | 100 | 0.77 | 0.023 |
| Chi2, region based | 76 | 0.73 | 0.005 |
| Chi2, case based | 76 | 0.75 | 0.015 |
| PCA, region based | 83 | 0.75 | 0.011 |
| PCA, case based | 83 | 0.79 | 0.041 |
| RPA, region based | 80 | 0.78 | 0.035 |
| RPA, case based | 80 | 0.84 | --- |