Optimal filtering strategies for task-specific functional PET imaging

Area of Science:

• Neuroimaging
• Medical Physics
• Biomedical Engineering

Background:

• Functional Positron Emission Tomography (fPET) is crucial for investigating brain function and diseases by analyzing dynamic metabolic and neurotransmitter processes.
• Optimizing signal processing in fPET is essential for accurately extracting task-specific information and improving diagnostic capabilities.

Purpose of the Study:

• To systematically evaluate and compare state-of-the-art filtering techniques for fPET imaging.
• To assess the impact of different filters on key performance metrics including reliability, signal-to-noise ratio, and spatial activation.
• To determine the optimal filter for reducing sample size requirements in fPET studies.

Main Methods:

• Forty healthy participants underwent [18F]FDG PET/MR scans while performing a cognitive task (Tetris®).
• Seven filtering techniques (3D/4D Gaussian, HYPR, IHYPR4D, MRI-MRF, dNLM, edNLM) were applied and evaluated using various hyperparameters.
• Performance was assessed via test-retest reliability, temporal signal-to-noise ratio (tSNR), spatial activation mapping, and sample size calculations.

Main Results:

• The extended dynamic Non-Local Means (edNLM) filter, along with dNLM, MRI-MRF, and IHYPR4D, significantly improved tSNR compared to 3D Gaussian smoothing.
• edNLM and HYPR filters notably enhanced test-retest reliability.
• NLM filters and MRI-MRF approaches improved spatial task-based activation, with edNLM reducing the required sample size by 15.4%.

Conclusions:

• The extended dynamic Non-Local Means (edNLM) filter exhibits promising performance across multiple critical metrics for fPET analysis.
• Filter selection in fPET imaging should be tailored to specific research objectives and available resources, considering trade-offs in performance.
• Optimized filtering techniques can enhance the efficiency and reliability of fPET studies, potentially reducing participant numbers and improving insights into brain dynamics.