Quantitative accuracy of MR based attenuation correction for PET - first experience with a whole-body PET/MR system

Ziel/Aim:

Combined whole body PET/MR systems have become commercially available recently. In 2010, one of the first of these systems (Philips Ingenuity TF PET/MR) has been installed at our institute. PET/MR is expected to offer new possibilities, in particular in the field of quantitative bimodal functional imaging (1). Quantitative PET image reconstruction requires attenuation correction (AC) which is commonly based on a direct measurement of photon attenuation using either a transmission scan with a Ge-68 radioactive source in standalone PET (TRAC) or a CT scan in PET/CT systems (CTAC). In PET/MR systems such a measurement cannot be performed. Hence, AC is performed with a software-based approach (MRAC) using dedicated tissue segmentation and tissue type identification (air, lung, soft tissue) of a MR image (2). Here, we report on a first evaluation of the accuracy of the manufacturer provided MRAC in whole body investigations with the Ingenuity TF PET/MR.

Methodik/Methods:

An evaluation of MRAC was performed by a direct comparison of the MR derived attenuation maps with transmission based attenuation maps acquired with a Siemens ECAT Exact HR+ PET system for 10 patients (8m/2f). We compared the pairs of coregistered attenuation maps in a voxel by voxel correlation analysis in the lung and the torso. In order to assess the influence of differences between the two AC methods on the resulting emission images we developed methods allowing to reconstruct PET emission data acquired at the PET/MR using the transmission based attenuation maps from the HR+ PET .

Ergebnisse/Results:

Our correlation analysis has shown that approximately 80% of the voxels in the lung and torso are in quantitative agreement. In one case, the MRAC algorithm failed to correctly detect the lung of the patient. In a further investigation, we observed metal artifacts resulting in distorted MR-derived attenuation maps.

Schlussfolgerungen/Conclusions:

The vendor-provided MRAC algorithm generally yields satisfactory results with respect to soft tissue and air segmentation. However, the algorithm relies on anatomic reference data and thus artifacts can arise if the anatomy of the patient does not fit to the reference. Additionally, metal artifacts can lead to distortion of the MR based attenuation maps.

Literatur/References:

(1) Pichler, B.J. et al., Sem. Nucl. Med., 38:199-208, 2008
(2) Martinez-Möller, A. et al., J. Nucl. Med., 50:520-526, 2009