Classification of parkinsonian syndromes via factorial discriminant analysis of brain SPECT data

Introduction. - The objective was to assess the value of single photon emission computerized tomography (SPECT) and factorial discriminant analysis (FDA) in the differential diagnosis of Parkinson's disease (PD), progressive supranuclear palsy (PSP), and corticobasal degeneration (CBD). Patients and methods. - Sixty-two patients with clinical diagnoses of either CBD, PSP or PD were studied using brain HmPaO-SPECT. Thirteen pairs of regions of interest (ROIs) were drawn on the slices located 50 mm and 90 mm above the canthomeatal plane. Twenty-six uptake indices and 13 asymmetry indices were determined. FDA was performed in order to determine whether or not the patients could be classified into the correct clinical group on the basis of SPECT data alone. The most discriminant parameters were used to generate two predictive scores, which were tested in a second group of 15 patients. Results. - FDA of all 39 variables correctly classified all the patients. A subset of 10 variables was used to build predictive scores, which correctly classified 90% of PD patients, 100% of PSP patients and 86% of CBD patients. When tested in the validation group of 15 patients, these predictive scores correctly classified 87% of the individuals. The frontal medial, temporoparietal and parietal regions were the most discriminant. Conclusion. - Using SPECT data alone, this study enabled us to distinguish between PD, PSP and CBD in patients with clear clinical presentations of the diseases in question. This novel, statistical approach provides reliable information. However, a prospective study dealing with de novo parkinsonian syndromes will be necessary. (C) 2008 Elsevier Masson SAS. All rights reserved.