Comparison of PET/CT post-reconstruction algorithms

Objectives The critical function of translating detected ionizing radiation emitted from patients into the formation of a representative map of physiological function is carried out by post-reconstruction algorithms. Accuracy of this rendering is paramount for both early detection and staging of lesions in the human body. The purpose of this study is to compare the image quality of a Hanning filtered-back-projection (FBP) to a Butterworth FBP algorithm. Three measurements were used to judge quality; Target Signal-to-Noise-Ratio (SNR), Background SNR and Contrast-to-Noise Ratio (CNR). Three fusion levels were evaluated.

Methods A NEMA IEC Body Phantom was acquired using a PET/CT scanner. The phantom consists of 6 sphere inserts with diameters of 10 mm, 13 mm, 17 mm, 22 mm, 28 mm, and 37 mm. Each sphere was filled with 18FDG and Ultravist 300 CT contrast using a 10:1 ratio. The acquisition was obtained at a typical clinical level using a 10:1 sphere-to-background ratio. The emission data was alternately reconstructed with Butterworth FBP and then with Hanning FBP. Utilizing manufacturer software, regions of interest were drawn for each sphere as well as their corresponding backgrounds. Mean and standard deviation information was recorded for each algorithm at the following levels of fusion: 100% CT, 50% CT/ 50% PET, and 100% PET. Target SNR, Background SNR and CNR were then calculated for each level.

Results At each fusion level, the Hanning FBP demonstrated superior Target SNR, Background SNR and CNR over the Butterworth FBP. Data from the 10 mm sphere was omitted due to lack of registration at each level of fusion for both Hanning FBP and Butterworth FBP.

Conclusions Hanning FBP provides a more realistic representation of lesion size than Butterworth FBP while minimizing the level of noise. Implementation of this post-reconstruction algorithm would improve the diagnosis and staging of lesions in the human body