Transmission Source Enhanced Attenuation Correction for High-Performance Brain PET

• Main Authors: Park, M.-A., Badawi, R. D., Bowen, S. L.
• Format: Conference Proceeding
• Language: English
• Subjects: Attenuation; Maximum likelihood detection; Microwave integrated circuits; Neuroimaging; Phantoms; Reconstruction algorithms; Semiconductor detectors
• ISSN: 2577-0829
• DOI: 10.1109/NSSMICRTSD49126.2023.10338087

Attenuation correction (AC) using the PET data alone is a promising option for accurate uptake quantification in standalone brain PET systems. However, current deep learning and joint reconstruction methods frequently exceed 5% uptake bias. Our objective is to optimize and evaluate an AC scheme for PET brain imaging that uses coincidences from a low-activity transmission source and the patient to reconstruct μ-maps from physics principles alone. We fabricated a sparse transmission source from a fillable tube (ID=1.6 mm) wrapped helically around the inside of a PET/CT bore. Our transmission aided maximum likelihood reconstruction of attenuation and activity (sTX-MLAA) includes a μ-map update that maximizes a log-likelihood of terms representing counts originating from all sources (i.e. external source and patient) and a hyperparameter weighted expression of counts segmented from the external source. Scatter is fully corrected from PET data alone. We evaluated performance of sTX-MLAA with a striatal brain phantom study on a time-of-flight PET/CT. The phantom was filled with 64 Cu (12.7 hr half-life) to approximate patient uptake of a dopamine transporter tracer, the transmission source filled with 18 F (1.8 hr half-life), and the phantom imaged for 30 min repeatedly over ~3 half-lives of 18 F; enabling an evaluation of the impact of transmission activity on performance with relatively fixed subject uptake. Qualitatively, PET images generated with AC from sTX-MLAA had high agreement with CT-AC data. With as little as 5 MBq in the transmission source, bias in tracer uptake was