Research Article

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

Table 1

Classifier performance for the different data sets (patients versus healthy controls, number of cases in brackets) in the LOOCV, without feature preselection. The column Perf. (%) indicates the percentage of subject cases correctly classified per group, Sensitivity (%) indicates the percentage of correctly classified healthy controls, and Specificity (%) indicates the percentage of correctly classified patients.

Feature set (size) Perf. (%) Sensitivity (%) Specificity (%)

PD-HC (38) 47.4 50 45
MSA-HC (39) 71.8 83.3 61.9
PSP-HC (35) 80.0 77.8 82.4