Energy-Efficient Seizure Detection Suitable for low-power Applications

TL;DR

The paper tackles the challenge of real-time seizure detection under stringent energy constraints for implantable and wearable devices. It introduces a compact TC-ResNet4 CNN trained without feature extraction, complemented by time-series post-processing methods (SMA, EWMA, HMM with Viterbi decoding) to boost detection while staying hardware-friendly. Validation on the CHB-MIT scalp EEG dataset demonstrates strong performance, achieving 95.28% accuracy, 92.34% sensitivity, and an AUC of 0.9384 with 4-bit fixed-point quantization, and confirms feasibility on the UltraTrail low-power accelerator with an average power of 495 nW. The work highlights a practical path to energy-efficient, real-time seizure detection suitable for low-power edge devices and neural implants, combining model compression with hardware-aware inference strategies.

Abstract

Epilepsy is the most common, chronic, neurological disease worldwide and is typically accompanied by reoccurring seizures. Neuro implants can be used for effective treatment by suppressing an upcoming seizure upon detection. Due to the restricted size and limited battery lifetime of those medical devices, the employed approach also needs to be limited in size and have low energy requirements. We present an energy-efficient seizure detection approach involving a TC-ResNet and time-series analysis which is suitable for low-power edge devices. The presented approach allows for accurate seizure detection without preceding feature extraction while considering the stringent hardware requirements of neural implants. The approach is validated using the CHB-MIT Scalp EEG Database with a 32-bit floating point model and a hardware suitable 4-bit fixed point model. The presented method achieves an accuracy of 95.28%, a sensitivity of 92.34% and an AUC score of 0.9384 on this dataset with 4-bit fixed point representation. Furthermore, the power consumption of the model is measured with the low-power AI accelerator UltraTrail, which only requires 495 nW on average. Due to this low-power consumption this classification approach is suitable for real-time seizure detection on low-power wearable devices such as neural implants.

Figures (7)

• Figure 1: Overview of the classification approach.
• Figure 2: Preprocessing pipeline.
• Figure 3: AUC scores computed for the base model, the retrained model and the retrained model with subsequent Viterbi decoding while the weights of the model were quantized to different bits.
• Figure 4: AUC scores computed for each patient and for each of the three time-series analysis method: SMA, EWMA and HMM.
• Figure 5: Metrics shown for each patient and for each of the three approaches in combination with the 4-bit quantized TC-ResNet4.