Dynamical Embedding of Single-Channel Electroencephalogram for Artifact Subspace Reconstruction
Doli Hazarika1, K N Vishnu1, Ramdas Ransing2
1Neural Engineering Lab, Department of Biosciences and Bioengineering, Indian Institute of Technology, Guwahati 781039, India.
Sensors (Basel, Switzerland)
|October 26, 2024
Summary
This study presents Embedded Artifact Subspace Reconstruction (E-ASR), a new method for cleaning single-channel electroencephalogram (EEG) data. E-ASR effectively removes artifacts, making mobile EEG monitoring more feasible.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Artifact Subspace Reconstruction (ASR) is effective for cleaning electroencephalogram (EEG) data but typically requires multiple channels.
- Single-channel EEG acquisition is desirable for portable and user-friendly applications, but artifact removal remains a challenge.
Purpose of the Study:
- To develop and evaluate a novel framework, Embedded ASR (E-ASR), for applying ASR to single-channel EEG data.
- To assess the efficacy of E-ASR in artifact reduction and signal reconstruction using quantitative metrics.
Main Methods:
- A dynamical embedding approach was used to create an embedded matrix from single-channel EEG data via delay vectors.
- The ASR algorithm was applied to the embedded matrix, followed by signal reconstruction.
- The E-ASR algorithm was tested on semi-simulated and real EEG data collected from four subjects using Fp1 and Fp2 electrodes.
Main Results:
- E-ASR achieved a relative root mean square error (RRMSE) of 43.87% and a correlation coefficient (CC) of 0.91 on semi-simulated data.
- The method effectively reduced artifacts in real EEG data, with validated eye-blink counts.
- Performance was evaluated using RRMSE, CC, and average power ratio.
Conclusions:
- The developed E-ASR framework successfully adapts the ASR algorithm for single-channel EEG data.
- This approach shows significant potential for artifact reduction in mobile EEG applications, enabling monitoring in naturalistic settings with minimal electrodes.
