Highly Efficient Compression Algorithms for Multichannel EEG.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Data Science

Background:

• High-dimensional electroencephalogram (EEG) data presents significant processing and understanding challenges.
• Efficient and robust lossless compression algorithms are crucial for managing large EEG datasets.

Purpose of the Study:

• To examine and compare various lossless compression techniques for single and multichannel EEG data.
• To propose novel, high-compression algorithms for multichannel EEG data, including general Multivariate Autoregression (MVAR) and modified context-based error modeling.

Main Methods:

• Evaluation of existing lossless compression methods: Huffman coding, arithmetic coding, Markov predictor, linear predictor, context-based error modeling, MVAR, and a bivariate model.
• Development and implementation of a general MVAR algorithm and a modified context-based error modeling approach for multichannel EEG.
• Validation using experimental multichannel EEG recordings and public standard databases.

Main Results:

• The proposed general MVAR and modified context-based error modeling achieved an average relative compression ratio of 70.64%, with some cases reaching 83.06%.
• The methods demonstrated superior performance compared to state-of-the-art compression techniques, as indicated by satisfactory parametric measures (percent-root-mean square distortion, peak signal-to-noise ratio, root-mean-square error, cross-correlation).

Conclusions:

• The developed compression algorithms efficiently reduce the storage and transmission bandwidth requirements for large multichannel EEG data.
• These advanced techniques offer a significant improvement over existing methods for EEG data compression, facilitating better data management and analysis.