Abstract
1775
Objectives Based on SSM/PCA toolbox, Parkinson’s disease Related Pattern (PDRP) was derived from FDG PET images which were usually reconstructed by filtered back projection (FBP) or ordered subsets expectation-maximization (OSEM). Since the use of SSM/PCA in multiple center clinical trial with different PET scanners and reconstruction algorithms, we used FDG PET scans of PD patients and age-matched healthy controls to confirm the influence of reconstruction algorithms.
Methods We used brain glucose metabolism scans of 58 PD patients and 22 controls.Every subject images were reconstructed by both FBP and OSEM with 3 iterations and 21 subsets. Subject PDRP scores were computed for images with both algorithms by SSM/PCA. Subject PDRP scores were correlated with the two algorithms. Group discrimination was also assessed by applying ROC analysis to PDs and controls. This analysis yielded values for the AUC and the sensitivity and specificity. In addition, subject scores for each algorithms of the PDRP were correlated with UPDRS motor ratings obtained for PD patients. These statistical computations were performed using SPSS software version 15 running under Windows XP. All comparisons were considered significant with P < 0.05.
Results PDRP scores for the two algorithms were significantly elevated in the PDs relative to the controls (p < 0.0001; t-test). PDRP scores from two algorithms were highly inter-correlated for the 58 patients (R = 0.998, p < 0.0001)and for the 22controls (R = 0.994, p < 0.0001). subject PDRP scores of FBP exhibited superior accuracy in the PDs from the 22 controls (AUC = 0.954, 95% CI=0.913-0.995; p<0.021), and subject PDRP scores of OSEM (AUC=0.951, 95% CI=0.908-0.993; p<0.022). PDRP scores of PDs by FBP(R = 0.592, p < 0.0001) and OSEM (R=0.585, p<0.0001) were both correlated with their UPDRS ).
Conclusions PDRP scores can be replicated in the different algorithms,FBP and OSEM. It suggested that SSM/PCA toolbox can be used in a wide variety of neurodegenerative disorders in multiple-center clinical trials.
Research Support National Nature Foundation of China (Grant No.81171189).