Attenuation Correction of the Cerebellum in PET/MR Data

Abstract

Attenuation correction (AC) is essential for achieving artefact-free PET/MR images. Many PET studies use the cerebellum as a reference region; therefore, AC methods should also be tested concerning their performance within the cerebellum. This study compares AC methods for PET/MR data, focusing on the cerebellum. Sixteen subjects underwent an [18F]FDG scan in a 3T-MR-BrainPET insert and a whole-head CT scan on the same day. The CT scan data were transformed into individual CT-based attenuation maps (AMCT), while the MR images were used to derive attenuation maps (AMs) using three methods: 1) Boston-MGH (AM textsubscript MGH); 2) London-UCL (AM textsubscript UCL); and 3) Juelich-Tx-template-based (AM textsubscript Tx-Juel). After reconstruction of the PET data with these four AMs, correlations, coefficients of determination, and relative errors (RErrs) between the PET-AM textsubscript CT and the other three PET-AMs were computed. The cerebellar RErr varied strongly between the three AC methods. The ${\mathrm{ AM}}_{MGH}$ method gave a RErr value of 3.85±5.03%, the AMUCL method gave 6.00±4.54%, and the AMTx-Juelgave 0.25±5.01%. Our results demonstrate that radiotracer uptake quantification in the cerebellum is sensitive to the applied PET AC. This dependency should be especially considered in neuroreceptor studies where the cerebellum is the reference region.