Low-Power Hardware Implementation of a Support Vector Machine Training and Classification for Neural Seizure Detection

Design of a hybrid AI network circuit for epilepsy detection with 97.5% accuracy and low cost-latency

Epilepsy detection using artificial intelligence (AI) networks has gained significant attention. However, existing methods face challenges in accuracy, computational cost, and speed. CNN excel in feature extraction but suffer from high computational latency and power consumption, while SVM rely heavily on feature quality and expensive kernel computations, limiting real-time performance. Additionally, most CNN-SVM hybrid model lack hardware optimization, leading to inefficient implementations with poor accuracy-latency trade-offs. To address these issues, this paper designs a hybrid AI network-based method for epilepsy detection using electroencephalography (EEG) signals. First, a hybrid AI network was constructed using three convolutional layers, three pooling layers, and a Gaussian kernel SVM to achieve EEG epilepsy detection. Then, the design of the multiply-accumulate circuit was completed using a parallel-style row computation method, and a pipelined convolutional computation circuit was used to accelerate the convolutional computation and reduce the computational overhead and delay. Finally, a single-precision floating-point exponential and logarithmic computation circuit was designed to improve the speed and accuracy of data computation. The digital back-end of the hardware circuit was realized under the TSMC 65 nm process. Experimental results show that the circuit occupies an area of 3.20 mm2, consumes 4.28 mW of power, operates at a frequency of 10 MHz, and has an epilepsy detection latency of 0.008 s, which represents a 32% reduction in latency compared to those reported in the relevant literature. The database test results showed an epilepsy detection accuracy of 97.5%, a sensitivity of 97.6%, and a specificity of 97.2%.

Seizure detection via reservoir computing in MoS2-based charge trap memory devices

Neurological disorders are a substantial global health burden, affecting millions of people worldwide. A key challenge in developing effective treatments and preventive measures is the realization of low-power wearable systems with early detection capabilities. Traditional strategies rely on machine learning algorithms, but their computational demands often exceed what miniaturized systems can provide. Neuromorphic computing, inspired by the human brain, demonstrated capabilities of on-chip computing with low power consumption. In this context, bidimensional (2D) semiconductors hold notable promise, thanks to their unique electronic properties, atomic-scale thickness, and scalability, making them ideal for low-power applications. This work presents a neuromorphic reservoir computing system exploiting MoS2-based charge trap memories (CTMs) for processing of electrophysiological signals. Real-time seizures detection is achieved, thanks to the nonlinear integration of local-field potential (LFP) recorded from in vitro rodent models of ictogenesis. The results support MoS2-based CTMs for low-power biomedical devices in clinical diagnosis and treatment of epilepsy.

Automated Age-Related Macular Degeneration Detector on Optical Coherence Tomography Images Using Slice-Sum Local Binary Patterns and Support Vector Machine

Artificial intelligence has revolutionised smart medicine, resulting in enhanced medical care. This study presents an automated detector chip for age-related macular degeneration (AMD) using a support vector machine (SVM) and three-dimensional (3D) optical coherence tomography (OCT) volume. The aim is to assist ophthalmologists by reducing the time-consuming AMD medical examination. Using the property of 3D OCT volume, a modified feature vector connected method called slice-sum is proposed, reducing computational complexity while maintaining high detection accuracy. Compared to previous methods, this method significantly reduces computational complexity by at least a hundredfold. Image adjustment and noise removal steps are excluded for classification accuracy, and the feature extraction algorithm of local binary patterns is determined based on hardware consumption considerations. Through optimisation of the feature vector connection method after feature extraction, the computational complexity of SVM detection is significantly reduced, making it applicable to similar 3D datasets. Additionally, the design supports model replacement, allowing users to train and update classification models as needed. Using TSMC 40 nm CMOS technology, the proposed detector achieves a core area of 0.12 mm2 while demonstrating a classification throughput of 8.87 decisions/s at a maximum operating frequency of 454.54 MHz. The detector achieves a final testing classification accuracy of 92.31%.

Auditory perception architecture with spiking neural network and implementation on FPGA

Spike-based perception brings up a new research idea in the field of neuromorphic engineering. A high-performance biologically inspired flexible spiking neural network (SNN) architecture provides a novel method for the exploration of perception mechanisms and the development of neuromorphic computing systems . In this article, we present a biological-inspired spike-based SNN perception digital system that can realize robust perception. The system employs a fully paralleled pipeline scheme to improve the performance and accelerate the processing of feature extraction. An auditory perception system prototype is realized on ten Intel Cyclone field-programmable gate arrays, which can reach the maximum frequency of 107.28 MHz and the maximum throughput of 5364 Mbps. Our design also achieves the power of 5. 148 W/system and energy efficiency of 845.85 μJ. Our auditory perception implementation is also proved to have superior robustness compared with other SNN systems. We use TIMIT digit speech in noise in accuracy testing. Result shows that it achieves up to 85.75% speech recognition accuracy under obvious noise conditions (signal-to-noise ratio of 20 dB) and maintain small accuracy attenuation with the decline of the signal-to-noise ratio. The overall performance of our proposed system outperforms the state-of-the-art perception system on SNN.

A medium-weight deep convolutional neural network-based approach for onset epileptic seizures classification in EEG signals

INTRODUCTION

Epileptic condition can be detected in EEG data seconds before it occurs, according to evidence. To overcome the related long-term mortality and morbidity from epileptic seizures, it is critical to make an initial diagnosis, uncover underlying causes, and avoid applicable risk factors. Progress in diagnosing onset epileptic seizures can ensure that seizures and destroyed damages are detectable at the time of manifestation. Previous seizure detection models had problems with the presence of multiple features, the lack of an appropriate signal descriptor, and the time-consuming analysis, all of which led to uncertainty and different interpretations. Deep learning has recently made tremendous progress in categorizing and detecting epilepsy.

METHOD

This work proposes an effective classification strategy in response to these issues. The discrete wavelet transform (DWT) is used to breakdown the EEG signal, and a deep convolutional neural network (DCNN) is used to diagnose epileptic seizures in the first phase. Using a medium-weight DCNN (mw-DCNN) architecture, we use a preprocess phase to improve the decision-maker method. The proposed approach was tested on the CHEG-MIT Scalp EEG database's collected EEG signals.

RESULT

The results of the studies reveal that the mw-DCNN algorithm produces proper classification results under various conditions. To solve the uncertainty challenge, K-fold cross-validation was used to assess the algorithm's repeatability at the test level, and the accuracies were evaluated in the range of 99%-100%.

CONCLUSION

The suggested structure can assist medical specialistsin analyzing epileptic seizures' EEG signals more precisely.