A hardware-algorithm co-design approach to optimize seizure detection algorithms for implantable applications.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Medical Devices

Background:

• Implantable neural prostheses offer a potential therapy for medically refractory epilepsy.
• Current seizure detection algorithms often require significant computational resources, limiting their use in implantable devices.
• Evaluating algorithms solely on detection efficacy overlooks critical hardware constraints like power consumption.

Purpose of the Study:

• To develop a two-dimensional design optimization approach for evaluating seizure detection algorithms for implantable neuroprosthetic devices.
• To consider both detection efficacy and hardware cost (power consumption) in algorithm feasibility analysis.
• To create a benchmark for comparing and selecting algorithms for next-generation epilepsy treatments.

Main Methods:

• Compared detection features for electrographic seizure detection using micro-electrode data from kainate-treated rats.
• Utilized circuit models to estimate dynamic and leakage power consumption for each feature.
• Developed a scoring system based on detection efficacy and hardware cost, plotted on a 2D design space.

Main Results:

• Identified an optimal combination of features maximizing detection efficacy per unit hardware cost.
• Demonstrated a method to quantitatively assess algorithm feasibility for battery-powered implantable applications.
• Established a framework for benchmarking seizure detection algorithms.

Conclusions:

• The presented two-dimensional optimization approach effectively balances seizure detection performance with hardware constraints.
• This method facilitates the development of efficient and effective implantable neuroprosthetic devices for epilepsy treatment.
• The findings support the creation of a common platform for evaluating seizure detection algorithms in future neuroprosthetic research.