Table 5 Additional imaging response detection performance metrics (Matthews correlation coefficients) for each model and the two types of classifier
From: High-dimensional detection of imaging response to treatment in multiple sclerosis
Predictor set | Support vector machines | Extremely randomised trees |
|---|---|---|
Brain volume | 0.132 [0.115–0.148] | 0.179 [0.164–0.194] |
Number of lesions | 0.168 [0.152–0.184] | 0.229 [0.214–0.243] |
Total lesion volume | 0.229 [0.213–0.245] | 0.239 [0.223–0.255] |
Best low-dimensional | 0.354 [0.339–0.369] | 0.355 [0.341–0.369] |
Regional atrophy | 0.576 [0.563–0.590] | 0.620 [0.607–0.633] |
Regional disconnection | 0.544 [0.530–0.559] | 0.559 [0.545–0.573] |
Best high-dimensional | 0.666 [0.653–0.679] | 0.675 [0.662–0.689] |
