MR-Based PET Attenuation Correction using 3D UTE/Multi-Echo Dixon: In Vivo Results

Background: PET/MR imaging offers promising opportunities for multi-parametric imaging that yields perfectly registered anatomical, functional and molecular information. Despite the growing interest and commercial success of PET/MRI scanners, PET attenuation correction remains a challenge for quantitative PET/MR imaging. Conventional MR-based methods often fail to provide continuous attenuation coefficients (AC) maps of short-T₂ components, i.e., bones and lungs, which account for the inter- and intra-subject variations. Recently, our group developed an MR-based method that utilizes a novel 3D UTE/Multi-echo Dixon sequence to generate continuous AC maps of lungs, bones, fat and soft tissues in a single acquisition [1]. In this study, we validated our method via in vivo human subject experiments.

Methods: Six volunteers were scanned under a study protocol that was approved by our local IRB. MR acquisitions were performed on a 3T MR scanner (Trio, Siemens Healthcare, Erlangen, Germany) using the proposed 3D UTE/Multi-echo Dixon sequence [1] with the following imaging parameters: image size = 128×128×128, resolution = 1.875×1.875×1.875 mm³, TE = 70, 2110, 2310, 3550, 4750, 4990, 5190 μs, and acquisition time = ~1 min (8x acceleration by compressed sensing). PET/CT imaging was performed on a PET/CT scanner (Discovery, GE Healthcare, Milwaukee, Wisconsin, USA). The imaging protocol of PET consisted of 15 mCi bolus injection of 11C PiB followed by a dynamic scan acquired over 70 min. Attenuation CT imaging was performed with tube peak voltage = 120 kVp and tube current time product 30 = mAs. The CT images were reconstructed for image size = 512x512x89, in-plane resolution = 0.59×0.59 mm², and slice-thickness = 3.75 mm. The linear attenuation coefficients (LAC) map was generated using the proposed method similar to previous work [1]. The proposed method was compared with the standard two-point Dixon method with CT LAC as the ground truth using Pearson’s correlation analysis.

Results: The proposed method allowed to estimate the continuous distribution of bone LAC (Fig.1A). The LAC map generated from the proposed method was visually comparable to those from CT (Fig.1A), showing higher correlation (r=0.95±0.02) compared to those from the two-point Dixon method (r=0.89±0.04). Attenuation correction using the LAC map from the two-point Dixon method resulted in an underestimation of PET signal, whereas the attenuation correction using the LAC map from the proposed method resulted in visually similar intensity compared to those reconstructed using the LAC map from CT (Fig.1B). Minor imperfection in registration was shown between MR and CT images, which is presumed to be due to approximately a month difference in MR and CT imaging, inter-scan motion, and the difference in the imaging system.

Conclusions: The performance of the MR-based AC method that utilizes 3D UTE/Multi-echo Dixon sequence has been validated via in vivo experiments. The proposed method allows to estimate the continuous distribution of bone LAC and can be potentially used to generate subject-specific LAC maps for PET/MR applications. Research Support: T32EB013180, R01CA165221, R01HL118261, R21EB021710, and P41EB022544 References: 1. Han PK, Horng DE, Lai X, El Fakhri G, Ouyang J, Ma C. MR-Based PET Attenuation Correction Using a 3D UTE/Multi-Echo Dixon Sequence. In Proceedings of the Annual Meeting of SNMMI, Philadelphia, USA, 2018, p.652. Figure 1. A: Representative result of linear attenuation coefficient (LAC) maps from CT, two-point Dixon, and the proposed method. B: Representative result of PET images reconstructed using LAC maps from CT, two-point Dixon, and the proposed method.

• Download figure
• Open in new tab
• Download powerpoint