Skip Navigation

Motion Correction of PET Images

Author: Nivedita Raghunath, MS

Overview: Patient head motion during a PET scan significantly degrades image quality, especially in high resolution scanners such as the High Resolution Research Tomograph (HRRT). At the Emory PET Center we have been routinely using the Vicra optical tracking system (Northern Digital Inc, Ontario, Canada) to track patient head motion.  We have developed a deconvolution algorithm in IDL to correct for head motion and successfully demonstrated improvement in image quality in phantom and volunteer scans. Our method has the potential for routine clinical usage and can be easily extended to other imaging modalities. Point sources can be used as a quality control measure to determine how much head motion is acceptable and when motion correction is necessary. This would enable quick decision making in a clinical setting.

Software Application for Motion Correction

Screenshot of the application

Screenshot of the application developed in IDL for motion correction of PET

Case study: We have performed several phantom and volunteer studies with motion.  In each study, a motion-free scan was performed for a duration of 20 minutes (also known as the reference scan), followed by another 20 minute scan with intentional head motion.

Figure below shows a Homer phantom experiment performed on the HRRT. 1.56 mCi of FDG was mixed with 400cc of water and filled into the basal ganglia. 1.36 mCi of FDG was mixed with 1800cc of water and filled into the rest of the phantom. 3 Na-22 point sources were taped to the head of the phantom.

Homer phantom scanned with motion correction

Figure 2

Figure 4

Homer phantom scanned with no motion (a1,a2), average motion of 16 mm in 6 discrete head positions (b1,b2) and corrected using deconvolution (c1,c2). a1,b1,c1 and a2,b2,c2 correspond to different slices of the Homer phantom.

There is great improvement in image quality/contrast as evident in the basal ganglia (indicated by arrows in c1). The blurred point source in b2 has successfully been resolved to a single point source in the corrected image c2. Figure 3 shows the profiles of the point sources and clearly shows the decrease in spread of the point sources from the blurred image to the corrected image.

Project Team

Nivedita Raghunath, MS, Tracy Faber, PhD, John Votaw, PhD