Table 1.
Demographic and Clinical Data for Participants.
Figure 1.
Predictive Probabilities for the Gaussian Process Classifier discriminating ADHD and Controls.
The x-axis describes the probability with which each subject is predicted to be an ADHD patient (equal to 1- the probability of being a control).
Figure 2.
2-class Multivariate and Conventional Maps.
A. Multivariate discrimination weight map for ADHD vs. Controls (unthresholded). Gaussian Process Classification classified ADHD patients and healthy controls with 82.8% and 75.9% sensitivity, respectively; leading to an overall accuracy of 79.3%. Multivariate discrimination weight-map –intensity values illustrate the relative positive weight distributions (ADHD; orange) and negative weight distributions (controls; light blue). Within each colour code, the lighter colors (i.e., light orange-yellow, light blue) indicate strongest weights for the GPC analyses and for the conventional mass-univariate case-control comparison lighter colors indicate higher p-values of structural differences. B. Multivariate discrimination weight map (thresholded). The map only shows voxels with a weight value above 40% of the maximum weight value C). Conventional mass-univariate t-statistic map. Controls had increased grey matter relative to patients, thresholded at cluster-wise p<0.001 uncorrected. No areas showed increased grey matter in ADHD relative to controls.
Figure 3.
2-class multivariate weight maps.
A) Multivariate discrimination weight map for ADHD vs. Controls (unthresholded). Gaussian Process Classification classified ADHD patients and healthy controls with 82.8% and 75.9% sensitivity, respectively; leading to an overall accuracy of 79.3%. Multivariate discrimination weight-map –intensity values illustrate the relative positive weight distributions (ADHD; orange) and negative weight distributions (controls; blue). Within each colour code, the lighter colors (i.e., light orange-yellow, light blue) indicate strongest weights for the GPC analyses. B) Multivariate discrimination weight map for ADHD vs. non-ADHD (unthresholded). Gaussian Process Classification classified ADHD patients and non-ADHD with 79.3% and 77.1% sensitivity, respectively; leading to an overall accuracy of 78.2%. Multivariate discrimination weight-map–intensity values illustrate the relative positive weight distributions (ADHD; orange) and negative weight distributions (non-ADHD; violet). Within each colour code, the lighter colors (i.e., light orange-yellow, light violet) indicate strongest weights for the GPC analyses. C) Multivariate discrimination weight map for ADHD vs. ASD (unthresholded). Gaussian Process Classification classified ADHD patients and ASD patients with 93.1% and 68.4% sensitivity, respectively; leading to an overall accuracy of 80.8%. Multivariate discrimination weight-map–intensity values illustrate the relative positive weight distributions (ADHD; orange) and negative weight distributions (ASD; green). Within each colour code, the lighter colors (i.e., light orange-yellow, light green) indicate strongest weights for the GPC analyses.
Table 2.
Global volume group differences in ADHD and controls.
Table 3.
Reduced grey matter in ADHD relative to healthy control boys in the traditional VBM analysis.