Quantification accuracy of neuro-oncology PET data as a function of emission scan duration in PET/MR compared to...

To evaluate and compare the effect of reduced acquisition time, as a surrogate of injected activity, on the PET quantification accuracy in PET/CT and PET/MR imaging.

Twenty min ¹⁸F-FDG phantom measurements and 10min ¹⁸F-FET brain scans were acquired in a Biograph-True-Point-True-View PET/CT (n=8) and a Biograph mMR PET/MR (n=16). Listmode data were repeatedly split into frames of 1min to 10min length and reconstructed using two different reconstruction settings of a 3D-OSEM algorithm: with post-filtering ("OSEM"), and without post-filtering but with resolution recovery ("PSF"). Recovery coefficients (RC_max, RC_A50) and standard uptake values (SUV_max, SUV_A50) were evaluated.

RC_max (phantom) and SUV_max (patients) increased significantly when reducing the frame duration. Significantly lower deviations were observed for RC_A50 and SUV_A50, respectively, making them more appropriate to compare PET studies at different number of counts. No statistical significant differences were observed when using post-filtering and reducing the frame time to 4min (RC_A50, reference 20min, phantom) and to 3min (SUV_A50, reference 10min, patients).

For hybrid aminoacid brain imaging, frame duration (or injected activity) can potentially be reduced to 30% of the standard used in clinical routine without significant changes on the quantification accuracy of the PET images if adequate reconstruction settings and quantitative measures are used. Frame times below 4min in the NEMA phantom are not advisable to obtain quantitative and reproducible measures.