Aim: The surface coils of the Biograph mMR integrated PET/MR system were optimised for PET, but are otherwise unaccounted for. The patient table is still more massive than those of PET/CT devices. The goal was to assess those hardware effects on quantification, count statistics, image quality and scan time both with phantoms and in patients and to investigate their clinical relevance.
Patients, material, methods: PET phantom data were acquired with and without the patient table. Image noise was expressed as relative standard deviation and compared to a state-of-the-art PET/CT scanner. Protocols of the phantom/patient study regarding the surface coils were similar. Thoraces/abdomens of 11 patients were scanned with and without a coil (1 BP, 4 min). Mean uptake and standard deviation in a cubical VOI were derived and expressed as SUV.
Results: The patient table reduced the number of true coincidences (trues) by 19% (PET/MR) and by 11% (PET/CT). The scan duration for the mMR had to be increased by approximately 30% to achieve a noise level comparable to that of the PET/CT. Decreased SUVs with coil observed in the phantom were confirmed by the patient study. By removing the coil, the mean liver SUV increased by (6 ± 2)%. With (+3 ± 14)%, the average change was similar in lesions, but exceeded 20% in almost one fifth of them. The number of trues grew by (6 ± 1)% for the patients and by 7% for the phantom.
Conclusion: Due to the additional attenuation caused by MR hardware, PET scan durations would have to be increased compared to current PET/CTs to provide similar image noise levels. The effect of the coils is mostly in the order of statistical fluctuations. In tumour lesions, it is more pronounced and shows a larger variability. Therefore, coils should be included in the attenuation correction to ensure accurate quantification and thus comparability across PET/MR and PET/CT scanners and within patient populations.
Keywords: MR hardware; PET quantification; PET/MR; hybrid imaging; multimodal imaging.