Real-Time Multi-Channel Epileptic Seizure Detection Exploiting an Ultra-Low-Complexity Algorithm-Hardware Co-Design...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Epileptic seizures require timely detection for effective management.
• Existing seizure detection methods often face challenges with computational complexity and hardware constraints.
• Developing robust, real-time seizure detection systems is crucial for patient care and research.

Purpose of the Study:

• To present an automated, threshold-based, multi-channel epileptic seizure detection algorithm.
• To design an algorithm suitable for low-complexity hardware implementations.
• To enhance detection reliability and minimize false alarms through a novel multi-channel strategy.

Main Methods:

• Utilized two computationally simple time-domain features (power and amplitude variations).
• Employed patient-specific offline calibration using statistical analysis of inter-ictal and ictal periods.
• Implemented a multi-channel decision-making strategy for improved robustness.
• Validated the algorithm on EEG CHB-MIT and iEEG SWEC-ETHZ datasets.
• Assessed hardware feasibility through FPGA synthesis, incorporating Time-Division Multiplexing (TDM).

Main Results:

• Achieved approximately 98% accuracy and over 98% sensitivity on both datasets.
• Reported average detection latencies of 3.37 s (EEG) and 7.84 s (iEEG).
• Demonstrated performance comparable to or exceeding machine-learning-based approaches.
• FPGA synthesis confirmed minimal and scalable resource requirements.

Conclusions:

• The proposed algorithm offers accurate and sensitive real-time epileptic seizure detection.
• Its multi-channel approach enhances reliability and reduces false alarms.
• The algorithm is well-suited for resource-constrained hardware, such as FPGAs, for low-complexity implementations.