Wearable systems for the continuous and real-time monitoring of cardiovascular diseases are becoming widespread and valuable assets in diagnosis and therapy. A promising approach for real-time analysis of the electrocardiographic (ECG) signal and the detection of heart conditions, such as arrhythmia, is represented by the transformer machine learning model. Transformers are powerful models for the classification of time series, although efficient implementation in the wearable domain raises significant design challenges, to combine adequate accuracy and a suitable complexity. In this work, we present a tiny transformer model for the analysis of the ECG signal, requiring only 6k parameters and reaching 98.97% accuracy in the recognition of the 5 most common arrhythmia classes from the MIT-BIH Arrhythmia database, assessed considering 8-bit integer inference as required for efficient execution on low-power microcontroller-based devices. We explored an augmentation-based training approach for improving the robustness against electrode motion artifacts noise, resulting in a worst-case post-deployment performance assessment of 98.36% accuracy. Suitability for wearable monitoring solutions is finally demonstrated through efficient deployment on the parallel ultra-low-power GAP9 processor, where inference execution requires 4.28ms and 0.09mJ.

The Internet of Medical Things (IoMT) paradigm is becoming mainstream in multiple clinical trials and healthcare procedures. Cardiovascular diseases monitoring, usually involving electrocardiogram (ECG) traces analysis, is one of the most promising and high-impact applications. Nevertheless, to fully exploit the potential of IoMT in this domain, some steps forward are needed. First, the edge-computing paradigm must be added to the picture. A certain level of near-sensor processing has to be enabled, to improve the scalability, portability, reliability and responsiveness of the IoMT nodes. Second, novel, increasingly accurate data analysis algorithms, such as those based on artificial intelligence and Deep Learning, must be exploited. To reach these objectives, designers, and programmers of IoMT nodes, have to face challenging optimization tasks, in order to execute fairly complex computing tasks on low-power wearable and portable processing systems, with tight power and battery lifetime budgets. In this work, we explore the implementation of a cognitive data analysis algorithm, based on a convolutional neural network trained to classify ECG waveforms, on a resource-constrained microcontroller-based computing platform. To minimize power consumption, we add an adaptivity layer that dynamically manages the hardware and software configuration of the device to adapt it at runtime to the required operating mode. Our experimental results show that adapting the node setup to the workload at runtime can save up to 50% power consumption. Our optimized and quantized neural network reaches an accuracy value higher than 97% for arrhythmia disorders detection on MIT-BIH Arrhythmia dataset.

Spiking Neural Networks (SNNs) are energy-and performance-efficient tools that have been found to be very useful in AI applications at the edge. This paper introduces SYNtzulu, an SNN processing element designed to be used in low-cost and low-power FPGA devices for near-sensor data analysis. The system is equipped with a RISC-V subsystem responsible for controlling the input/output and setting runtime parameters, thus increasing its flexibility. We evaluated the system, which was implemented on a Lattice iCE40UP5K FPGA, in various use cases employing SNNs with accuracy comparable to the state-of-the-art. SYNtzulu dissipates a maximum power of 12.05 mW when performing SNN inference, which can be reduced to an average of just 1.45 mW through the use of dynamic power management.

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. Tests conducted on the CHB-MIT dataset show a 20% reduction of the onset detection latency with respect to the state-of-the-art model for temporal acquisition, with a competitive 73% seizure detection probability and 0.15 false-positive-per-hour (FP/h). Further investigations on a novel and challenging scalp EEG dataset result in the successful detection of 88% of the annotated seizure events, with 0.45 FP/h. We evaluate the deployment of the EEGformer on three commercial low-power computing platforms: the single-core Apollo4 MCU and the GAP8 and GAP9 parallel MCUs. The most efficient implementation (on GAP9) results in as low as 13.7 ms and 0.31 mJ per inference, demonstrating the feasibility of deploying the EEGformer on wearable seizure detection systems with reduced channel count and multi-day battery duration.

The accurate modeling of hand movement based on the analysis of surface electromyographic (sEMG) signals offers exciting opportunities for the development of complex prosthetic devices and human-machine interfaces, moving from discrete gesture recognition, towards continuous movement tracking. In this study, we present two solutions for real-time sEMG processing, based on lightweight Spiking Neural Networks (SNNs) and efficiently implemented on a Lattice iCE40-UltraPlus FPGA, especially suitable for low-power applications. We first assess the performance in the discrete finger gesture recognition task, considering as a reference the NinaPro DB5 dataset, and demonstrating an accuracy of 83.17% in the classification of twelve different finger gestures. We also consider the more challenging problem of continuous finger force modeling, referencing the Hyser dataset for finger tracking during independent extension and contraction exercises. The assessment reveals a correlation of up to 0.875 with the ground-truth forces. Our systems take advantage of SNNs’ inherent efficiency and, dissipating 11.31 mW in active mode, consume 44.6 μJ for a gesture recognition classification and 1.19 μJ for a force modeling inference. Considering dynamic power-consumption management and the introduction of idle periods, average power drops to 1.84 mW and 3.69 mW for these respective tasks.