Beads phantom for evaluating heterogeneity of SUV on ¹⁸F-FDG PET images

This study aimed to develop a dedicated phantom using acrylic beads for texture analysis and to represent heterogeneous ¹⁸F-fluorodeoxyglucose (FDG) distributions in various acquisition periods.

Images of acrylic spherical beads with or without diameters of 5- and 10-mm representing heterogeneous and homogeneous ¹⁸F-FDG distribution in phantoms, respectively, were collected for 20 min in list mode. Phantom data were reconstructed using three-dimensional ordered subset expectation maximization with attenuation and scatter corrections, and the time-of-flight algorithm. The beads phantom images were acquired twice to evaluate the robustness of texture features. Thirty-one texture features were extracted, and the robustness of texture feature values was evaluated by calculating the percentage of coefficient of variation (%COV) and intraclass coefficient of correlation (ICC). Cross-correlation coefficients among texture feature values were clustered to classify the characteristics of these features.

Heterogeneous ¹⁸F-FDG distribution was represented by the beads phantom images. The agreements of %COV between two measurements were acceptable (ICC ≥ 0.71). All texture features were classified into four groups. Among 31 texture features, 24 exhibited significant different values between phantoms with and without beads in 1-, 2-, 3-, 4-, 5-, 20-min image acquisitions. Whereas, the homogeneous and heterogeneous ¹⁸F-FDG distribution could not be discriminated by seven texture features: low gray-level run emphasis, high gray-level run emphasis, short-run low gray-level emphasis, low gray-level zone emphasis, high gray-level zone emphasis, short-zone low gray-level emphasis, and coarseness.

We have developed the acrylic beads phantom for texture analysis that could represent heterogeneous ¹⁸F-FDG distributions in various acquisition periods. Most texture features could discriminate homogeneous and heterogeneous ¹⁸F-FDG distributions in the beads phantom images.