Balancing accuracy, delay and battery autonomy for pervasive seizure detection.

Area of Science:

• Neuroscience and Biomedical Engineering
• Epilepsy Research and Treatment Technologies

Background:

• Closed-loop neurostimulation presents a promising approach for treating absence seizures by detecting and suppressing epileptic activity.
• Existing devices often compromise on detection speed, accuracy, or energy efficiency, limiting their suitability for continuous operation.
• There is a critical need for algorithms that optimize these parameters for practical, low-power implantable or wearable seizure detection systems.

Purpose of the Study:

• To explore the design space of a novel seizure-detection algorithm for absence seizures.
• To evaluate the trade-offs between detection accuracy, speed, and device autonomy on a low-power processor.
• To develop an algorithm suitable for continuous closed-loop neurostimulation.

Main Methods:

• Development and evaluation of a novel seizure-detection algorithm employing a complex Morlet wavelet filter and static thresholding.
• In-depth design space exploration focusing on accuracy, detection speed, and energy consumption.
• Analysis of algorithm performance on a low-power processor to assess device autonomy.

Main Results:

• A minimal reduction in average detection rate (1.83%) significantly enhances device autonomy (3.7x).
• The optimized algorithm achieves faster seizure detection times (from 710 ms to 540 ms).
• Demonstrated feasibility of balancing detection performance with energy constraints for wearable/implantable devices.

Conclusions:

• The novel seizure-detection algorithm effectively balances accuracy, speed, and energy efficiency for absence seizures.
• This approach enables substantial improvements in device autonomy and detection speed, crucial for closed-loop neurostimulation.
• The findings pave the way for more practical and effective wearable or implantable epilepsy management solutions.