Paola Busia, Andrea Cossettini, Thorir M Ingolfsson, Simone Benatti, Alessio Burrello, Victor J B Jung, Moritz Scherer, Matteo A Scrugli, Adriano Bernini, Pauline Ducouret, Philippe Ryvlin, Paolo Meloni, Luca Benini
The long-term, continuous analysis of electroencephalography (EEG) signals on wearable devices to automatically detect seizures in epileptic patients is a high-potential application field for deep neural networks, and specifically for transformers, which are highly suited for end-to-end time series processing without handcrafted feature extraction. In this work, we propose a small-scale transformer detector, the EEGformer, compatible with unobtrusive acquisition setups that use only the temporal channels. EEGformer is the result of a hardware-oriented design exploration, aiming for efficient execution on tiny low-power micro-controller units (MCUs) and low latency and false alarm rate to increase patient and caregiver acceptance...
January 23, 2024: IEEE Transactions on Biomedical Circuits and Systems