Pulsation artifact removal from intra-operatively recorded local field potentials using sparse signal processing and...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Neural recordings, particularly local field potentials (LFPs), are often corrupted by high-amplitude artifacts from electrocardiogram (ECG) or pulsation.
• These artifacts obscure neural patterns, hindering accurate analysis and visual inspection of brain activity during intraoperative recordings.

Purpose of the Study:

• To develop and evaluate a novel sparse signal representation strategy for denoising pulsation artifacts in intraoperative LFP recordings.
• To enhance the clarity and reliability of neural data by effectively removing confounding artifactual components.

Main Methods:

• A sparse signal representation approach using K-singular value decomposition (K-SVD) to create a data-specific dictionary.
• Artifact morphology estimation anchored by QRS-peaks detected from ECG.
• Orthogonal matching pursuit (OMP) to reconstruct artifact components and residual calculation for denoising.

Main Results:

• Significant reduction in pulsation artifact amplitude in LFP signals.
• Improved visualization of neural oscillations in alpha and beta bands.
• Noticeable signal strength recovery in low-frequency bands (<13 Hz) previously masked by artifacts.
• Substantial increase in the signal-to-noise ratio (SNR) of the denoised LFP data.

Conclusions:

• Sparse signal representation offers an effective method for denoising pulsation artifacts in LFP recordings.
• The proposed technique reconstructs quasi-periodic pulsation templates, enabling the computation of artifact-free neural activity.
• This approach improves the quality of intraoperative neural recordings, facilitating better clinical interpretation.