Using EQ·PET to reduce reconstruction-dependent variations in [¹⁸F]FDG-PET brain imaging

Area of Science:

• Nuclear Medicine
• Medical Imaging Analysis

Background:

• Positron Emission Tomography (PET) imaging, particularly with 2-deoxy-2[18F]fluoro-D-glucose ([18F]FDG), is crucial for diagnosing neurological conditions like early-onset Alzheimer's disease (EOAD).
• Variations in image reconstruction protocols across different PET scanners can lead to quantification differences, hindering reliable comparisons of [18F]FDG PET data.

Purpose of the Study:

• To assess the efficacy of the European Association of Nuclear Medicine (EANM) harmonisation strategy combined with the EQ·PET methodology for harmonizing brain [18F]FDG PET images.
• To evaluate the ability of EQ·PET to minimize reconstruction-induced variability in [18F]FDG PET quantification for early-onset Alzheimer's disease patients.

Main Methods:

• Utilized the NEMA NU 2 body phantom with [18F]FDG and reconstructed data using three common clinical protocols: OSEM 3D with Time of Flight (TOF) at 2 and 6 iterations, and OSEM 3D with TOF and Point Spread Function (PSF) at 8 iterations.
• Computed EQ·PET filters as Gaussian smoothing to align recovery coefficients (RCs) from different reconstructions to a reference.
• Applied the EQ·PET filter to clinical [18F]FDG PET brain images from 35 EOAD patients, assessing performance through qualitative and quantitative metrics on the cortical surface, with and without partial volume correction.

Main Results:

• The EQ·PET methodology successfully identified optimal smoothing to minimize root-mean-square error (RMSE), harmonizing brain [18F]FDG PET images and enabling reliable neurological comparisons.
• Performance was superior for TOF reconstructions compared to TOF + PSF reconstructions.
• While EQ·PET significantly reduced reconstruction-induced variability, moderate differences persisted between harmonized PSF and standard OSEM reconstructions, indicating a need for caution with PSF modeling.

Conclusions:

• The EQ·PET methodology, integrated with the EANM harmonisation strategy, effectively minimizes reconstruction-induced variability in brain [18F]FDG PET imaging.
• This approach enhances the comparability of [18F]FDG PET data for neurological applications, particularly in EOAD.
• Careful consideration is advised when harmonizing PET images reconstructed with PSF modeling due to residual variability compared to standard OSEM methods.