The content discusses the development of a real-time classification system for electroneurographic (ENG) signals to support the recovery of patients with peripheral nerve injuries. Key highlights:
Peripheral neuropathies are a significant clinical challenge, and the use of fully implanted devices is a promising solution. However, these devices pose numerous challenges, including the real-time classification of motor/sensory stimuli from the ENG signal.
The authors propose a multiple-input multiple-output (MIMO) ENG signal model that accounts for the aggregate propagation of nerve motor and sensory signals. This model is used to design and evaluate different artificial neural network (ANN) architectures for real-time ENG signal classification.
Four ANN architectures are investigated: Convolutional Neural Networks (CNNs), Inception Time (IT), Electroencephalogram Network (ENGNet), and Long Short-Term Memory (LSTM) networks. These networks are designed to extract spatial-temporal features from the ENG signals and classify them in real-time.
The performance of the ANN-based classifiers is evaluated on real ENG data sets measured from the sciatic nerves of rats. The results show that some ANNs can achieve accuracies over 90% for signal windows of 100 and 200 ms, with low enough processing time to be effective for real-time applications.
The analysis provides insights into the effectiveness and efficiency of the proposed architecture and modeling in different conditions, highlighting the boundaries of real-time ENG signal classification.
To Another Language
from source content
arxiv.org
Thông tin chi tiết chính được chắt lọc từ
by ntonio Covie... lúc arxiv.org 04-01-2024
https://arxiv.org/pdf/2403.20234.pdfYêu cầu sâu hơn