Unencrypted login | home

Program Information

Compressed Sensing in PET Imaging with Partial Detector Rings

no image available
S Valiollahzadeh

S Valiollahzadeh1*, T Chang1, J Clark1, O Mawlawi2, (1) Rice University, Houston, TX (2) MD Anderson Cancer Center, Houston, TX

WE-C-217BCD-9 Wednesday 10:30:00 AM - 12:30:00 PM Room: 217BCD

Purpose: Investigate the possibility of reducing the number of PET detector elements per ring (introduce gaps) while maintaining image quality by employing compressed sensing techniques.

Methods: A uniform Ge-68 phantom was imaged on a D-RX PET/CT scanner twice; once with all detectors operational (baseline) and once with 8 equidistant detector blocks turned off (partially sampled [PS], 11% detectors off. The resulting PS sinogram was then decomposed in two different components, each sparsely represented in a specific transform domain. An iterative optimization technique was then used to recover the PS sinogram based on the solution of a combination of underdetermined system of equations and block-coordinated relaxations. In addition, the total variation is minimized for the first component to direct it into a piece-wise smooth model. Finally the two components were summed to obtain the sinogram which was used to compare with the original PS sinogram. Comparison was done only for existing sinogram pixel values. This process was repeated iteratively until a RMS error of 5% or a total of 100 iterations were reached. For each iteration update, the values of the pixels corresponding to the missing detectors were obtained from the previous iteration while the remaining pixel values were extracted from the baseline sinogram. The resultant corrected sinograms where then reconstructed using OSEM and FBP and the corresponding images for full and PS sinograms were compared using mean and max activity concentration in a ROI placed centrally over the phantom.

Results: For OSEM (FBP) reconstruction, the mean and max activity concentration difference were -0.06% (-0.02%) and 3.7% (5.5%) respectively when compared to baseline.

Conclusions: Compressed sensing seems to have the ability to recover PS PET data. Such an approach can potentially be used to generate PET images with accurate quantitation while reducing number of detectors/ring.

Contact Email