PET Attenuation Correction using Synthetic CT from Ultrashort Echo-time MRI

Jump to: navigation, search

PET Attenuation Correction using Synthetic CT from Ultrashort Echo-time MRI

Snehashis Roy, Wen-Tung Wang, Aaron Carass, Jerry L. Prince, John A. Butman, and Dzung L. Pham


Integrated PET (Positron Emission Tomography)/MR (magnetic resonance) systems are becoming more popular in clinical and research applications. Quantitative PET reconstruction requires correction for γ photon attenuations using an attenuation coefficient map ( µ-map), that is a measure of the electron density. One challenge of PET/MR, in contrast to PET/CT is the accurate computation of µ-maps. Unlike CT, MRI measures physical properties not directly related to electron density. Thus, the attenuation coefficients are usually computed using either segmentation of the MR images or using deformable registration.


PET Fig1.png
Figure 1: Top two rows show dual echo UTE images (TE=70µs and 2.46ms) and corresponding original CT based µ-maps of a reference and a subject with a lesion. Bottom row shows Siemens Dixon [5], Siemens UTE based µ-map and our GENESIS result for the subject.

In this work, we propose an example-based method to generate whole head µ-maps from Ultra-short Echo Time (UTE) MR imaging sequences. UTE images are preferred to other MR sequences, because signal from bone is present in the scan with ultra short echo (~100 µs). To generate a synthetic CT image we use patches from a reference dataset, which consists of dual echo UTE images and a co-registered CT from the same subject. Patches of the UTE images under study are matched to relevant patches in the reference UTE images so that corresponding patches from the CT images can be combined via a Bayesian framework. No registration or segmentation is required.


For evaluation, UTE, CT, and PET data was acquired from five patients under an IRB approved protocol. We used another patient (with UTE and CT only) as a reference to generate synthetic CT images for the five. PET reconstructions were attenuation corrected using (1) the original CT, (2) our synthetic CT, and (3) segmentation based synthetic µ-maps, which are Siemens Dixon and UTE based µ-maps and (4) a deformable registration based CT. Our synthetic CT based PET reconstruction shows higher correlation (average p = 0.99, R^2= 0.99) to the original CT based PET, as compared to the segmentation and registration based methods. Synthetic CT based reconstruction had minimal bias (regression slope 0.99) as compared to the segmentation based methods (regression slope 0.97). A peak signal-to-noise ratio of 35.98 dB in the reconstructed PET activity is observed, compared with 29.67 dB for Dixon and Siemens UTE based µ-maps.

PET Fig4.png
Figure 4: (a)-(b) shows UTE images of a subject, (c)-(g) shows Siemens UTE and Dixon µ-maps along with the ones generated from a registration based method [12], GENESIS, and original CT, (h)-(l) shows the reconstructed PET images using the corresponding µ-maps, (m)-(p) shows the absolute difference between original CT based PET and PET reconstructed from other four µ-maps. The linear colormap ranges from 0 to 25000 for PET images and 0 to 2500 for difference images.


A patch-matching approach to synthesize CT images from dual echo UTE images leads to significantly more accurate PET reconstruction as compared to actual CT scans. The PET reconstruction is improved over Dixon and Siemens scanner generated u-map based results, even in subjects with pathology.


Matlab executables are available.


  • S. Roy, A. Carass, A. Jog,and J.L. Prince, "MR to CT registration of brains using image synthesis", Proceedings of SPIE Medical Imaging (SPIE-MI 2014), San Diego, CA, February 15-20, 2014.
  • S. Roy, A. Jog, A. Carass, and J.L. Prince, "Atlas Based Intensity Transformation of Brain MR Images", MBIA, 51-62, 2013.


  • W.T. Dixon, "Simple proton spectroscopic imaging", Radiology, 153: 189-194, 1984.
  • N. Burgos, M.J. Cardoso, M. Modat et. al., "Attenuation Correction Synthesis for Hybrid PET-MR Scanners", 16th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2013), Nagoya, Japan, September 22-26, 2013.