BrainFuseNet: Enhancing Wearable Seizure Detection Through EEG-PPG-Accelerometer Sensor Fusion and Efficient Edge Deployment
Thorir Mar Ingolfsson, Xiaying Wang, Upasana Chakraborty, Simone Benatti, Adriano Bernini, Pauline Ducouret, Philippe Ryvlin, Sándor Beniczky, Luca Benini, Andrea Cossettini
Abstract
This paper introduces <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">BrainFuseNet</small>, a novel lightweight seizure detection network based on the sensor fusion of electroencephalography (EEG) with photoplethysmography (PPG) and accelerometer (ACC) signals, tailored for low-channel count wearable systems. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">BrainFuseNet</small> utilizes the Sensitivity-Specificity Weighted Cross-Entropy (SSWCE), an innovative loss function incorporating sensitivity and specificity, to address the challenge of heavily unbalanced datasets. The <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">BrainFuseNet</small>-SSWCE approach successfully detects <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$93.5\%$</tex-math></inline-formula> seizure events on the CHB-MIT dataset (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$76.34\%$</tex-math></inline-formula> sample-based sensitivity), for EEG-based classification with only four channels. On the PEDESITE dataset, we demonstrate a sample-based sensitivity and false positive rate of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$60.66\%$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$1.18$</tex-math></inline-formula> FP/h, respectively, when considering EEG data alone. Additionally, we demonstrate that integrating PPG signals increases the sensitivity to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$61.22\%$</tex-math></inline-formula> (successfully detecting <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$92\%$</tex-math></inline-formula> seizure events) while decreasing the number of false positives to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$1.0$</tex-math></inline-formula> FP/h. Finally, when ACC data are also considered, the sensitivity increases to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$64.28\%$</tex-math></inline-formula> (successfully detecting <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$95\%$</tex-math></inline-formula> seizure events) and the number of false positives drops to only <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$0.21$</tex-math></inline-formula> FP/h for sample-based estimations, with less than one false alarm per day when considering event-based estimations. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">BrainFuseNet</small> is resource-friendly and well-suited for implementation on low-power embedded platforms, and we evaluate its performance on GAP9, a state-of-the-art parallel ultra-low power (PULP) microcontroller for tiny Machine Learning applications on wearables. The implementation on GAP9 achieves an energy efficiency of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$21.43$</tex-math></inline-formula> GMAC/s/W, with an energy consumption per inference of only <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$0.11$</tex-math></inline-formula> mJ at high performance (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$412.54$</tex-math></inline-formula> MMAC/s). The <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">BrainFuseNet</small>-SSWCE method demonstrates effective and accurate seizure detection on heavily imbalanced datasets while achieving state-of-the-art performance in the false positive rate and being well-suited for deployment on energy-constrained edge devices.