Figure 1 | Scientific Reports

Figure 1

From: Brain multi-contrast, multi-atlas segmentation of diffusion tensor imaging and ensemble learning automatically diagnose late-life depression

Figure 1

Shown in descending order are the twenty highest trace (a) and RD (b) relative feature importance determined as the normalized weighted sum of the number of times a given feature is used to split the data in the ensemble. The higher the frequency of a feature being used for splitting, the higher its importance. For example if the fornix has a feature importance of 0.2 it means fornix has a relative importance of 0.2 or a 20% in the ensemble (averaged over 30 repetitions of classification) compared to the other features. Gm gray matter.

Back to article page