Reliability of drawing regions of interest for positron emission tomographic data.

We determined interrater reliability for two raters who independently used a standard protocol to draw regions of interest (ROIs) on F-18 fluorodeoxyglucose positron emission tomographic data acquired from eight patients with mild to moderate memory impairment. Intraclass correlation coefficients for region sizes (total pixel number) varied among anatomical ROIs (RIs from 0.324 to 0.935). However, correlations calculated from relative metabolic rates (normalized average counts) were consistently high (RI > or = 0.954). The raters also showed high agreement for region recognition on each image plan (kappa = 0.978). These results suggest that rater disagreements in ROI margins have minimal impact on average count densities used to calculate metabolic rates.