Towards MR-based attenuation correction for whole-body PET/MR imaging

Abstract

1469

Objectives: The success of whole-body (WB) PET/CT has led to an increased interest also in combined PET/MR as a complementary hardware combination. However, neither standard nor X-ray transmission sources can be installed easily in a PET/MRT. Therefore, the original MR images should be used for attenuation correction (AC) of the emission data. We present a first approach to MR-based AC for WB-PET/MR studies. In the absence of WB-PET/MR tomographs we propose a software-based approach for MR-based AC.

Methods: In this ongoing study we present initial results from 100 consecutive WB-patients imaged on PET/CT and MR on the same day. PET/CT scans were acquired on a Biograph Classic with a limited breath-hold protocol (370MBq FDG, 1h pi, 3.5min/bed, CT: 130kVp, 110mAs, 5mm slices, CT-AC). WB-MR scans (T1w, T2w) were acquired pre- and post Gd-contrast with an 1.5T system using a rolling table platform covering 7 axial scan positions. Thorax-MR images were acquired in normal inspiration. The MR images were used to generate pseudo-CT images following a histogram-matching intensity transformation, which excluded all pixels with gray values < mean of all pixel values in the MR volume (ITK user guide). These pseudo-CT images were co-registered to the CT images of the WB-PET/CT study using a commercially available, non-linear registration algorithm (Fusion7D, Siemens) before being exported to the PET/CT and used for AC. PET image reconstruction was performed using AW-OSEM (2iter, 8subs) on 128-matrices. Using circular ROI’s we estimate tumor-to-muscle (T/M) as well as lung-to-blood (L/B) ratios on the corrected PET images from the PET/CT and pseudo-PET/MR study.

Results: Figure 1 shows a patient with a lesion near the lumbar spine. T/M and L/B in PET/CT following CT-AC were 2.0 and 0.3, respectively. Following MR-based AC, T/M and L/B were 1.7 and 0.3, respectively. T/M differences could be caused by the different position of the arms. We are currently working on a refinement of the co-registration step, whereby we intent to use a validated, non-rigid registration algorithm. Validation for thoracic regions will be performed separately using a set of spirometer-triggered 3D-thoracic MR data sets of volunteers.

Conclusions: MR-based AC of whole-body PET studies is feasible. However, refinement of the intensity transformation for various clinical MR sequences is required, and dedicated bone sequences should be included.