Dynamic dictionary for combined EEG compression and seizure detection

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Real-time electroencephalogram (EEG) data requires efficient compression for storage and transmission.
• Existing compression techniques may not adequately preserve diagnostic information, particularly for seizure detection.

Purpose of the Study:

• To develop a novel, real-time EEG compression technique.
• To evaluate the method's efficacy in both data compression and seizure-like activity detection.

Main Methods:

• Utilized discrete wavelet transform (DWT) to exploit redundancy across frequency subbands.
• Employed dynamic reference lists for decorrelating subband coefficients.
• Applied Set Partitioning in Hierarchical Trees (SPIHT) as a source coder.

Main Results:

• Achieved significant compression of one-dimensional (1-D) EEG channels.
• Demonstrated successful detection of seizure-like activity at bitrates as low as 2 bits per sample.
• Diagnostics-oriented assessment confirmed both compression and detection capabilities.

Conclusions:

• The proposed method offers effective real-time EEG compression.
• The technique shows promise for integrated seizure detection within the compression algorithm.
• This approach can maintain diagnostic utility at low bitrates.