Classification of Parkinsonian Syndromes from FDG-PET Brain Data Using Decision Trees with SSM/PCA Features

Table 2

Classifier performance with preselection of features (patients versus healthy controls, number of cases in brackets). The percentage of principal components is arranged in order of highest to lowest variance accounted for and best number of PCs according to AIC. Highest performances are in bold.

%/number of PCs	In order of amount of variance					According to AIC
	3%	5%	50%	70%	100%	1	3	5
PD-HC (38)	55.3	63.2	57.9	63.2	47.4	63.2	50	47.4
MSA-HC (39)	71.8	74.4	69.2	71.8	71.8	66.7	69.2	74.4
PSP-HC (35)	82.9	80	77.1	77.1	80	82.9	80	80