Previous studies on PET quantification phantoms have shown that the presence of a non-active wall decreases the lesion-to-background contrast. This creates inaccuracies in quantification, particularly when imaging a small sphere where the wall represents a large portion of its overall volume. The presence of cold walls has also been shown to affect the measured standard uptake values, which reduce as wall thickness increases.

To address these shortcomings, Marie Sydoff and colleagues at Lund University's Department of Medical Radiation Physics in Malmö have developed a simple method for creating radionuclide-doped gelatin phantoms without non-active walls. To compare these with conventional phantoms, they used a Jaszczak phantom containing three hollow plastic spheres (with inner diameters of 15.6, 22.5 and 27.9 mm) in a polymethylmethacrylate (PMMA) cylinder.

The researchers dissolved 18F-FDG in water to an activity concentration of approximately 80 kBq/ml, simulating a clinical situation. They added gelatin granules to half of the solution and water to the other half. The gelatin solution was heated, poured into spherical aluminium moulds (with diameters of 15.6, 22.5 and 27.9 mm) and then frozen to create solid spheres. The other half of the solution was used to fill the hollow plastic spheres. Finally, all six spheres were mounted in the Jaszczak phantom.

The spheres were imaged with a Gemini TF PET/CT system using a matrix size of 144 × 144, a voxel size of 4 × 4 × 4 mm, and a scan time of 8 min/bed position. Measurements were made with the spheres surrounded by water (zero background), and with background fractions of 0.08, 0.1, 0.13 and 0.2.

After PET image reconstruction and attenuation correction, the data were processed to calculate the background-corrected relative threshold – Tvol – for all of the spheres. Tvol provides an intensity threshold with which to delineate the volumetric edges of a tumour or spherical phantom. For the largest spheres and the lowest background fraction (0.08), Tvol was lower for the hollow plastic sphere (39.7%) than for the gelatin sphere (41.4%). With the highest background (0.2), Tvol was 33.7% for the large plastic sphere and 41.1% for the large gelatin sphere.

The researchers plotted Tvol as a function of the five background fractions. With zero background activity, the Tvol values for the gelatin and hollow spheres were practically the same. As the background activity increased, however, Tvol for all of the hollow spheres decreased while Tvol for the gelatin spheres remained practically constant.

"When we used the Tvol values for volume calculations, the values obtained from the hollow spheres showed increasing overestimation of volume with increasing background activity levels," explained Sydoff. "This makes the Tvol values for the gelatin spheres more accurate for structures in an active background, which is always the case in patient measurements."

The authors conclude that threshold values estimated using spheres with non-active walls should not be used for tumour delineation in patients, especially for small volumes within a high-activity background. With the gelatin spheres, however, the background-corrected threshold method can be employed to define tumour volume from PET images.

The researchers now plan to study the effects of plastic walls in micro-PET systems, where phantom walls are a much bigger problem, since the volumes are far smaller. "I am also looking at the possibilities of using this method of volume estimation on structures other than spheres, to be able to implement it in real patient measurements such as within oncology, for example," said Sydoff.

Related stories