Skip to main content
. Author manuscript; available in PMC: 2016 Nov 1.
Published in final edited form as: Int J Dev Neurosci. 2015 Aug 24;46:125–131. doi: 10.1016/j.ijdevneu.2015.07.007

Table 2.

Ten features having the greatest contribution to classification accuracy in the SVM

Overall Rank Average Rank Hemisphere Average regional thickness (SD), mm p Region
Non-converted Converted
1 1.00 right 2.64 (0.18) 2.45 (0.15) 0.004 Medial orbitofrontal cortex
2 2.52 right 2.58 (0.22) 2.64 (0.10) 0.351 Precentral gyrus
3 3.66 left 2.84 (0.20) 2.95 (0.19) 0.137 Rostral anterior cingulate cortex
4 5.22 left 2.95 (0.12) 3.09 (0.19) 0.015 Insula
5 6.10 right 2.97 (0.19) 2.95 (0.22) 0.770 Insula
6 8.38 left 2.53 (0.13) 2.57 (0.13) 0.312 Lateral orbital frontal cortex
7 8.52 right 2.75 (0.25) 2.87 (0.20) 0.156 Rostral anterior cingulate cortex
8 9.80 right 2.72 (0.13) 2.74 (0.18) 0.648 Inferior temporal gyrus
9 10.94 right 2.64 (0.18) 2.66 (0.16) 0.717 Superior temporal gyrus
10 11.14 right 2.60 (0.22) 2.48 (0.12) 0.051 Lateral orbital frontal cortex

Laterality, region name, and average feature weight rank of the highest ranked features are displayed. Regions having a lower overall rank contributed a greater amount of information to the classifier. Significance values indicate uncorrected p values from post-hoc comparisons of regional thickness averages between groups using two-tailed independent samples t-tests.