TABLE 4

Agreement of Face and Face Patch Classification With Clinical Categorization

HC Versus FASHC Versus FAS + PFAS
CMLDASVMCMLDASVM
Face0.9670.9671.000.8920.9090.909
Periorbit0.9830.9170.9670.8920.9000.892
Perioral0.8500.8500.8840.8830.8830.883
Perinasal0.8330.8500.9340.8250.8250.817
Profile0.9330.9330.9170.9250.9330.917
  • Each classification rate was estimated as the mean area under ROC curves of 20 cross-validation trials and corresponds to the probability of correctly classifying 2 individuals, 1 taken from each of the 2 groups being compared. Closest mean classification labels members of 2 groups by the name of the group whose mean is most similar. For linear discriminant analysis (LDA), the goal is a linear combination of principal component modes that exhibits the largest difference in the subgroup means relative to the within-group variance. Support vector machines, or large margin classifiers, focus on individual cases in the overlap of the subgroups to be classified that help to define a separating surface with largest margin between the subgroups. In addition to the full face, patches of the face were also considered in isolation: periorbit, perioral, perinasal, and profile. CM, closest mean; SVM, support vector machines.