Systematic analysis and comparison of commercial seizure-detection software

Area of Science:

• Neurology
• Medical Technology
• Epilepsy Research

Background:

• Epilepsy management relies on accurate seizure detection.
• Long-term video-electroencephalography monitoring (VEM) is crucial for diagnosis and treatment.
• Automated seizure detection software aims to improve efficiency and accuracy.

Purpose of the Study:

• To evaluate the performance of three commercial seizure-detection software packages: Besa 2.0, Encevis 1.7, and Persyst 13.
• To compare their sensitivity, specificity (false alarm rate), and detection delay in epilepsy patients undergoing VEM.
• To assess the accuracy of these tools in a clinical setting.

Main Methods:

• A comparative study involving 81 epilepsy patients undergoing long-term VEM.
• Analysis of seizure detection performance metrics: sensitivity, specificity, and detection delay.
• Statistical comparison of Besa 2.0, Encevis 1.7, and Persyst 13 using kappa statistics.

Main Results:

• No significant difference in per-patient seizure detection rates (sensitivity) among Besa (67.6%), Encevis (77.8%), and Persyst (81%).
• Significant differences in false alarm rates (per hour): Besa (0.7/h), Encevis (0.2/h), Persyst (0.9/h).
• Significant differences in detection delay (seconds): Besa (30s), Encevis (25s), Persyst (20s). Persyst 13 had the shortest delay.

Conclusions:

• Commercially available seizure-detection software packages offer comparable sensitivity for seizure detection.
• Significant variations exist in false alarm rates and detection delays among these software packages.
• Persyst 13 exhibited the highest detection rate and shortest delay, while Encevis 1.7 had the lowest false alarm rate.