Table 2.

Performance summary.

Dataset
Classifier	SVM			Adaboost
Dimensionality Reduction	SFS	LDA	Independent	SFS
Prewhitened
Functional	Acc. = 78% Sen. = 65% Spe. = 65%	Acc. = 65%	Acc. = 65%	Acc. = 63% Sen. = 44% Spe. = 85%
AAL	Acc. = 80% Sen. = 77% Spe. = 68%	Acc. = 67%	Acc. = 73%	Acc. = 62% Sen. = 52% Spe. = 80%
Raw
Functional	Acc. = 64%	Acc. = 63%		Acc. = 63%
AAL	Acc. = 64%	Acc. = 65%		Acc. = 61%

Classification accuracy, sensitivity and specificity for different datasets, classifiers, dimensionality reduction methods, and atlases. SFS, Sequential Forward Selection; LDA, Linear Discriminant Analysis; Independent, Selecting top features based on their independent performance. Since classification performance using raw data was lower than that of prewhitened data using SFS and LDA feature selection methods, classification performance on raw data using the independent feature selection method is not reported for either atlas. Bold values represent the highest value obtained for each measure.