Accurate attenuation correction despite movement during PET imaging

Abstract

Recent methods to realign PET images are useful to correct minor changes in position due to subject movement. When using measured attenuation correction in the presence of larger subject movements, the attenuation data will not correctly represent the conditions of the emission scan and hence lead to erroneous attenuation correction.

Methods: In the presence of subject movement between two emission scans (EM1 and EM2), the first scan was assumed to be correctly aligned with the transmission scan. The emission scans were reconstructed without attenuation correction and were realigned so that the transformation matrix mapping EM2 onto EM1 was found. The inverse of that matrix was used to reslice the transmission images. The resliced transmission images were forward projected to yield new attenuation data and EM2 was re-reconstructed. Finally, the image set from EM2 was resliced to match the orientation of EM1. The method was validated by phantom data and by human brain and cardiac studies.

Results: Phantom data show that accurate attenuation correction can be achieved by reslicing the transmission images. In 795 region of interest measurements, the mean error after reslicing was -0.8%, the standard deviation of the error was 3.8% and the maximum error was no greater than 11.7%. Results from human brain and cardiac experiments yield similar results with a maximum error of 7.5% in 80 measurements.

Conclusion: Reslicing of attenuation data with transformation parameters obtained from realignment of emission data provides an accurate and reliable method to obtain correct quantification in the presence of changes in position due to subject movement during or between PET examinations.