Combining time series and frequency domain analysis for a automatic seizure detection

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epileptic seizure detection in long-term electroencephalographic (EEG) recordings is challenging and requires expert analysis.
• Existing algorithms like EpiScan show high performance but can be further optimized.
• Irregular and distorted rhythmic activity in EEG poses a significant detection challenge.

Purpose of the Study:

• To develop and evaluate a novel time domain method for detecting epileptic seizure patterns, particularly irregular and distorted rhythmic activity.
• To enhance the performance of existing seizure detection systems by integrating the new method.
• To improve the accuracy and reduce false alarms in automated EEG-based seizure detection.

Main Methods:

• Development of a novel time domain algorithm focusing on irregular and distorted EEG rhythms.
• Scanning EEG data for sequences of similar epileptiform discharges.
• Utilizing a combination of duration and similarity measures for seizure classification.
• Integration and testing of the novel method with the established EpiScan algorithm.

Main Results:

• The novel method, when combined with EpiScan, increased overall seizure detection sensitivity from 70% to 73%.
• The combined approach reduced the false alarm rate from 0.33 to 0.30 alarms per hour.
• The method was validated on a large EEG database comprising 275 patients and over 22,000 hours of recordings.

Conclusions:

• The novel time domain method effectively detects epileptic seizure patterns, including challenging irregular and distorted activities.
• Integrating this method with EpiScan offers a significant improvement in automated seizure detection performance.
• This advancement has the potential to reduce the burden on medical experts and improve patient care through more accurate seizure identification.