Meher Rohit Juttukonda, Presenter: Research Grant, Siemens AG

Yasheng Chen PhD, Abstract Co-Author: Nothing to Disclose

Yueh Z. Lee MD, PhD, Abstract Co-Author: Research Grant, Carestream Health, Inc

Felipe Espinoza, Abstract Co-Author: Nothing to Disclose

Weili Lin PhD, Abstract Co-Author: Nothing to Disclose

Dinggang Shen PhD, Abstract Co-Author: Nothing to Disclose

David Lalush, Abstract Co-Author: Nothing to Disclose

Hongyu An DSc, Abstract Co-Author: Nothing to Disclose

Since MR signal depends on proton (not electron) density and tissue relaxation, MR-based attenuation correction (AC) for MR/PET imaging is challenging. In this study, we have developed an atlas-based MR method for PET AC and have evaluated its accuracy using simulations.

In a group of 10 patients with paired MR and CT images, 3 AC maps were generated for each patient. First, gold-standard maps (CTSC) were derived using bilinear scaling of their CT images. Second, nonlinear registration was performed to generate a T1w-MR and CT atlas. A hidden Markov random field method was then used in a leave-one-out scheme to derive a pseudo-CT (pCT) for each patient using the atlas and their MR image. These pCTs were then scaled to obtain MR-based AC maps (MRAC). Last, the original CTs were segmented into 3 classes (bone, air, soft tissue) and each class was assigned its attenuation value to yield a third AC map (CTSEG). The CTSEG method is considered the silver standard because it represents the best possible outcome of using an MR-segmentation-based AC method. Realistic, patient-specific PET simulations were then conducted to evaluate the performance of MRAC and CTSEG in the 10 patients for two cases: 1) normal brain and 2) brain with a focal lesion added. Attenuation was applied at the sinogram stage using the CTSC map and was corrected for using CTSEG and MRAC. The results were reconstructed and a percent-error analysis against the gold standard was performed.

The MRAC method resulted in a lower average whole-brain absolute percent-error (WBAE) across patients of 1.84% compared to the 4.57% achieved by the CTSEG method (p<0.001). In addition, the MRAC method also produced a narrower error distribution with an average full-width tenth-maximum (FWTM) of 6.96% compared to the 14.24% average FWTM produced by the CTSEG method (p<0.001). Finally, the WBAE for the MRAC method in the cases with the focal lesions (1.88%) was not significantly different from the normal cases (p=0.43).

We have developed a MR T1w-based AC method that outperforms the silver standard. In addition, the presence of a focal lesion does not affect the accuracy of the proposed method.

An accurate MR-based AC method allows the MR/PET to become quantitatively reliable and opens the door for simultaneous quantitative-PET and MR applications.

Juttukonda, M, Chen, Y, Lee, Y, Espinoza, F, Lin, W, Shen, D, Lalush, D, An, H, A Novel MR Atlas-based Method Outperforms the Silver-standard CT Segmentation-based Method in PET Attenuation Correction.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13025381.html