Ing. Kateřina Dudášová, Ph.D.

Neuroimaging of subtle hypometabolic epileptogenic lesions using 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) is challenging due to its limited spatial resolution, which causes partial volume effects (PVE), leading to image blurring and underestimation of radiotracer activity. Partial volume correction (PVC) post-processing compensates for the blurring, but relies on accurate knowledge of the point spread function (PSF), which must be provided as an input parameter. However, no guidelines exist for determining an appropriate PSF value, which varies with scanner type, reconstruction algorithm, and scanned object properties. We conducted phantom measurements on Siemens Biograph mCT 40 and Vision 600 scanners using a clinical reconstruction protocol to compare two PSF estimation methods: one based on capillary phantom (following NEMA) and another using effective image resolution (EIR) of Hoffman phantom images (following EARL guideline). The impact of the PSF estimation method on PVC performance was assessed using image quality metrics: gray matter recovery coefficient (GMRC), gray matter edge recovery coefficient (GMERC), and gray matter coefficient of variation (GMCOV). PVC with both PSF estimation methods improved image quality compared to the original images. However, PVC with input PSF estimated by EIR showed better performance than PVC with capillary-based estimate: GMRC was 92-94% vs. 85-86%, GMERC was 85-90% vs. 78-79%, and lower GMCOV: 18-22% vs. 20-23%. These findings suggest that the estimation of input PSF by EIR could be a more reliable and practical approach for PVC in clinical FDG-PET brain imaging, especially since it is already used in the EARL F/C Brain PET/CT Accreditation.