Gross target volume delineation on PET images by a numerical approximation method phantom studies

A scheme named SUV_-Shape for drawing the gross target volume (GTV) on PET images was developed by a numerical approximation method and evaluated by several simulating tumor models. Globe-shaped, ellipsoid-shaped, cavity-shaped, leaflet-shaped, club-shaped hollow spheres were made, and their median volume was 19.61 (5.88, 64.56) mL. These models filled with 23.72 kBq/mL∼162.03 kBq/mL [18F]-FDG representing a tumor and fixed in a barrel filled with 3.99 kBq/mL∼8.13 kBq/mL [18F]-FDG representing the background served as phantoms. And the tumor-background ratio ranged from 2.92∼ 40.6 in the four phantoms. [18F]-FDG 3D PET-CT images of phantoms at rest were acquired. SUV_-Shape scheme measured the GTV of the tumor by a numerical approximation method. The paired-t tests and Spearman's correlation coefficient (rho) between GTVs and the volume of models (GTVt) were done. The measurement error analyses were done for the GTVs. Results 27 tumor models were drawn by SUV_-Shape, and their median GTVs was 20.56 (5.1, 65.16) mL. In addition; these models' GTVt and GTVs were not significant different (t =-1.27, P = 0.22). Moreover, rho between them were 0.959 (P < 0.01), which suggested they were highly relative. According to the location error was 1.95 mm, 20 tumor models' GTVs were within their margin of relative error; while the reminder seven models' GTVs were beyond them. SUV_-Shape could delineate and measure the GTV of simulating tumor, and their position resolutions of sub-pixel were usually achieved. (authors)