EEG analysis via multiscale Lempel-Ziv complexity for seizure detection

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Seizure detection and prediction remain significant clinical challenges.
• Lempel-Ziv Complexity (LZC) is a recognized feature for seizure detection.
• Augmenting LZC can enhance analysis of biomedical signals by emphasizing amplitude or frequency variations.

Purpose of the Study:

• To evaluate the feasibility of Multiscale Lempel-Ziv Complexity (MLZC) for seizure detection.
• To compare the performance of MLZC against LZC in distinguishing ictal and non-ictal periods.

Main Methods:

• The study introduces and applies the Multiscale Lempel-Ziv Complexity (MLZC) feature.
• MLZC was compared with Lempel-Ziv Complexity (LZC) for seizure detection.
• Performance was evaluated across three distinct cases using patient seizure recordings.

Main Results:

• MLZC demonstrated a clear separation between non-ictal and ictal periods in all three evaluated cases.
• A single threshold was effective with MLZC across 7 recordings and 7 seizures per patient.
• LZC provided clear separation in only one of the three cases, indicating MLZC's superior performance.

Conclusions:

• MLZC is a feasible and effective feature for seizure detection.
• MLZC offers improved performance over LZC by preventing high-frequency components from being overwhelmed.
• The findings suggest MLZC's potential for more robust seizure detection systems.