Development and Validation of an Extended NEMA Body Phantom to Detect 18F-FDG Heterogeneous Distribution in Texture Analysis

We evaluated heterogeneous 18 F-fluorodeoxyglucose (FDG) distribution using texture features and an extended National Electrical Manufacturers Association (NEMA) phantom. Homogeneous and heterogeneous distribution was represented by spheres without or with acrylic beads (5-mm diameter). We reconstructed PET images acquired for 10 min using the Q.clear algorithm and controlled statistical noise by adjusting β values. We calculated maximum (SUV max ) and mean (SUV mean ) standardized uptake values and coefficients of variation (COV), then evaluated 22 established texture features to determine homogeneity and heterogeneity. Heterogeneous and homogeneous 18 F-FDG distribution was visualized using spherical phantoms with and without beads. The SUV max and SUV mean were underestimated by ~ 45% in heterogeneous compared with homogeneous 18 F-FDG distribution. As the β value increased, the COV for the phantoms with and without beads decreased and eventually became equivalent. When the spherical phantoms with or without beads were imaged using a β value of 100, eight textural features were associated with texture heterogeneity, and six features were associated with texture homogeneity. We extended a NEMA phantom by inserting spheres containing beads to represent heterogeneous 18 F-FDG distribution. Robust texture features were determined using the phantom to detect 18 F-FDG homogeneity and heterogeneity.