Feasibility of semiautomatic analysis on F-18 FP-CIT PET with a standardized template: Comparison with manual and MR-based methods

초록

Objectives The aim of this study was to evaluate the feasibility of FP-CIT template-based quantitative analysis on F-18 FP-CIT PET in patients with de novo Parkinson's disease, compared with manually-driven and semiautomatic MR-based methods. Methods Forty-five de novo PD patients (68.6±8.9y) underwent both MRI and F-18 FP-CIT PET. Images were spatially normalized to a standardized FP-CIT template containing 4 ROIs of putamen, caudate, occipital cortex and cerebellum. Mean counts were obtained using the quantification program, Korean Structural Statistical Probabilistic Anatomical Map (KSPAM) by Lee et al. Putamen-to-caudate ratio (PCR), specific-to-nonspecific ratio with 2 different references: to occpital; V3″(occ) and to cerebellum; V3″(cbll), were compared. Parameters were also calculated from manual method using surface-fitting technique, and MR-based method with an MR template. Spearman’s rho (ρ), bland-altman plot, concordance-correlation coefficient were determined. Results Means ± SD's from manual, MR and PET-based methods were respectively as follows: 0.93±0.2, 1.09±0.3, 1.49±0.5 for PCR, 1.78±0.6, 1.18±0.4, 1.29±0.6 for V3″(occ), and 2.31±0.7, 1.64±0.5, 1.33±0.5 for V3″(cbll). Manual ROI technique was highly reproducible (ICC=0.70-0.94). All parameters were significantly correlated across all methods (ρ=0.37-0.88, p< 0.01). V3″(occ) revealed better correlation than V3″(cbll) (ρ=0.53-0.79, p<0.001). Although CCC were variable, suggesting fair to moderate agreement between three methods, V3″(occ) (0.39-0.49) tended to be more concordant than V3″(cbll) (0.19-0.35). Conclusions FP-CIT template-based method showed good correlation with manual and MR-based methods. We also found a better agreement with a reference of occipital cortex rather than cerebellum. It suggests that FP-CIT template-based method may replace both laborious manual and MR-based method.