EEG compression using JPEG2000: how much loss is too much?

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Biosignal compression is crucial for power conservation in wireless body area networks and ambulatory monitoring.
• Lossy compression offers higher compression ratios and power savings than lossless methods but risks signal degradation.
• The diagnostic impact of lossy compression on Electroencephalogram (EEG) data requires further investigation.

Purpose of the Study:

• To apply a variant of the lossy JPEG2000 algorithm to EEG data.
• To assess the impact of varying compression parameters on EEG signal fidelity.
• To determine the acceptable level of signal degradation for clinical diagnostic information using a seizure detection algorithm.

Main Methods:

• A variant of the lossy JPEG2000 algorithm was applied to EEG data from the Freiburg epilepsy database.
• Compression parameters were varied to produce EEG reconstructions with different signal fidelities.
• The reconstructed EEG signals were analyzed using the REACT seizure detection algorithm to evaluate diagnostic performance.

Main Results:

• The study analyzed hundreds of hours of reconstructed EEG data to quantify the impact of lossy compression on seizure detection.
• It identified the extent of EEG signal distortion that can be tolerated without losing diagnostically significant information.
• Achievable compression ratios for different levels of signal fidelity were presented.

Conclusions:

• Lossy compression of EEG data using JPEG2000 can achieve significant power savings.
• The study provides insights into the trade-off between compression levels and diagnostic accuracy for epilepsy monitoring.
• This research helps establish guidelines for acceptable signal degradation in compressed EEG for automated seizure detection.