Research Article

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

Table 3

Performance for binary classification of disease groups in the LOOCV. The number of cases per group is in brackets. The column Perf. indicates the percentage of subject cases correctly classified (all features included), Sensitivity indicates the percentage of correctly classified first disease group, Specificity indicates the percentage of correctly classified second disease group, and Perf. (AIC-5) indicates the performance when features are reduced to the best 5 PCs according to AIC.

Group Perf. (%) Sensitivity SpecificityPerf. (AIC-5) (%)

PD versus MSA (41) 73.27076.278
PD versus PSP (37) 67.68052.970.3
MSA versus PSP (38) 68.476.258.871.1