High-temporal resolution metabolic connectivity resolved by component-based noise correction

Area of Science:

• Neuroimaging
• Molecular Imaging
• Biophysics

Background:

• Functional PET (fPET) offers second-level temporal resolution for metabolic process modeling.
• High-temporal fPET is sensitive to noise, challenging signal extraction for applications like molecular connectivity imaging.
• Existing preprocessing methods may not optimally handle the noise characteristics of high-temporal fPET data.

Purpose of the Study:

• To adapt and evaluate a component-based noise reduction strategy (CompCor) for high-temporal [18F]FDG-fPET scans.
• To assess the impact of different filtering methods and scanner types (PET/CT vs. PET/MR) on metabolic connectivity (M-MC) estimation.
• To compare the characteristics of fPET-derived networks with those from functional MRI (fMRI).

Main Methods:

• Applied a CompCor preprocessing pipeline, adapted from fMRI, to high-temporal [18F]FDG-fPET data from PET/CT (1s frames) and PET/MR (3s frames) scanners.
• Compared various filtering strategies across different frequency bands to assess their effect on M-MC.
• Analyzed the consistency, structure, and spatial characteristics of resulting connectivity networks.

Main Results:

• CompCor yielded more consistent and structured metabolic networks compared to standard bandpass filters.
• Intermediate frequency bands (0.01-0.1 Hz) demonstrated the most reliable M-MC across both PET/CT and PET/MR scanners (r=0.89).
• High-sensitivity PET/CT also showed structured patterns at higher frequencies (0.1-0.2 Hz). fPET networks were spatially cohesive but less differentiated than fMRI networks.

Conclusions:

• High-temporal [18F]FDG-fPET, when combined with appropriate denoising techniques like CompCor, allows for reliable estimation of resting-state metabolic connectivity.
• Scanner characteristics and preprocessing choices critically influence signal quality and network reliability.
• This physiologically informed pipeline enhances cross-system and cross-study comparability in molecular imaging.