Using MLAA and registration to estimate attenuation of a flexible body matrix coil in nonTOF PET/MRI

The attenuation of PET gamma photons caused by magnetic resonance (MR) hardware (HW) is often overlooked for flexible coils, potentially resulting in an underestimation of patient activity and reduced reproducibility. This study compares the application of a maximum likelihood algorithm that updates activity and attenuation using emission data only (MLAA) alone to its application combined with registration of a pre-acquired CT-derived template μMap of the coil on full count data as well as a sixth of the data to simulate gated PET data. NonTOF PET/MR scans of a patient were performed, where the MR body coil was placed on the patient torso. The data was then reconstructed using MLAA with known patient attenuation, but unknown HW attenuation. A template μMap was created from a CT-scan of the same coil wrapped around a NEMA Body phantom. This μMap was projected along the vertical axis, registered to the projected MLAA output and shifted to align with the patient outline in vertical direction. All μMaps were then used in final maximum likelihood expectation maximisation (MLEM) reconstructions of the PET data. This procedure was repeated for a sixth of the acquired counts to investigate the applicability of these methods on gated PET data. The highly attenuating preamplifiers were on the whole moved to the right position by registration and vertical shifting, although the coil deformation was not represented accurately by the current method. This remains the case when using a sixth of the data. When a background-prior was used, the step-size of MLAA had to be reduced to reach a converged μMap. The activity images reconstructed with the obtained HW μMaps show an increased activity uptake with respect to the reference scan with patient attenuation correction only.