Image Signal-To-Noise Equalization in Whole Body PET Using Variable Acquisition Times
M Dahlbom*, A Kriszan, J Czernin, UCLA School of Medicine, Los Angeles, CASU-E-I-82 Sunday 3:00:00 PM - 6:00:00 PM Room: Exhibit Hall
Whole Body PET scans are acquired at multiple axial positions, where the acquisition time at each position is constant. Although the acquisition time is adjusted for patient weight, the varying amount of attenuation and activity distribution for different sections, the image S/N can vary significantly. The aim of this work is to investigate the use of variable bed position scan times in WB-PET to equalize the Signal-to-Noise ratio in the axial direction.
Simulations of activity and attenuation distributions based on whole body CT scans were performed. Phantoms of different cross sections were also simulated and imaged. Image noise was estimated by generating multiple noise replicates by adding Poisson noise to the emission sinograms for the simulated images, and using a bootstrap method for the phantom patient measurements. By comparing the square of image noise (SD/Mean) for all the image slices, the acquisition time for each section could be adjusted to yield uniform image noise for all slices. The image noise was also compared to the average AC factors through the center of each body slice.
A polynomial function was found for both simulations and the measurements to accurately describe image noise as a function of AC factors. Using this relationship, the acquisition time at each axial position can be adjusted to produce images of relatively uniform S/N, independent of cross sectional thickness. This was confirmed in phantom and patient data.
The noise properties of WB-PET images can be equalized axially by adjusting the acquisition time according to the amount of attenuation. The acquisition time can be reduced in areas of lower attenuation and increased in more absorbing sections Since there is a correlation of the image noise and the CT-derived AC factors, the acquisition times can quickly calculated using a simple functional relationship.