Sebastian Robert McWilliams MBBCh, Presenter: Nothing to Disclose

Yi Su PhD, Abstract Co-Author: Nothing to Disclose

Richard Laforest PhD, Abstract Co-Author: Nothing to Disclose

Brian Rubin, Abstract Co-Author: Nothing to Disclose

Jonathan Edward McConathy MD, PhD, Abstract Co-Author: Research Consultant, GLG Consulting Speakers Bureau, Eli Lilly and Company Research Consultant, General Electric Company

Agus Priatna PhD, Abstract Co-Author: Research Consultant, Siemens AG Employee, Siemens AG

Tammie Smith Benzinger MD, PhD, Abstract Co-Author: Research Grant, Eli Lilly and Company

MR attenuation correction (MRAC) in hybrid PET/MR imaging poses challenges for accurate quantification. CT attenuation correction (CTAC) provides an approximation of mass attenuation based on tissue density. To assess the accuracy of factory MRAC methods, two methods were compared to CTAC for a series of patients that had PET/CT and PET/MR as part of the same cerebral amyloid PET study.

Study patients underwent 18F-florbetapir imaging on a Biograph mMR scanner with reference PET/CT images acquired on a Biograph 40 PET/CT scanner utilizing the same tracer injection; emission images used for analysis were acquired from 50-70 minutes post-injection. Dixon and UTE-based MRAC were performed on the PET/MR-acquired data and attenuation maps generated. The CTAC attenuation map (mu-map) was co-registered with the MRAC calculated maps. In order to analyze structure dependent variation, segmentation of the cerebral structures was performed using FreeSurfer on T1-weighted MR data and PET emission data. Standardized uptake value ratios (SUVRs) were calculated relative to the cerebellar cortex for MRAC and CTAC data and their correlation assessed with a linear best-fit line.

UTE MRAC incorrectly assigned air attenuation coefficients to intracranial structures thus was excluded from analysis. Compared to CTAC, there was a 9.6% underestimation of activities in all regions by Dixon MRAC. The inferior temporal cortex (-18.2%), pars orbitalis (-16.3%), and lateral occipital cortex (-14.4%) had the highest percent error. Accuracy improved for deeper structures including the precuneus (-2.6%), thalamus (+1.9%), and cerebellar cortex (-9.6%). The precuneus SUVR was overestimated by 7.8% and the mean cortical SUVR (MCSUVR) by 5.3%. The SUVR values were correlated at the precuneus (r=0.95) and for the MCSUVR (r=0.97).

When compared to CTAC-corrected activities and SUVRs, Dixon MRAC generally led to underestimation of values with spatial variation. There was strong correlation of SUVRs between the two methods. Further patients are being analyzed to identify the most accurate method of MRAC when applied to cerebral amyloid PET/MR imaging.

The optimum MR attenuation correction method for PET/MR is not established in cerebral amyloid PET. Comparison with coregistered CTAC images allows accuracy of different MRAC methods to be assessed.

McWilliams, S, Su, Y, Laforest, R, Rubin, B, McConathy, J, Priatna, A, Benzinger, T, Evaluating the Optimum MR Based Attenuation Correction Method for F-18 Florbetapir Imaging on Hybrid PET/MR.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13044318.html