基于注意力增强CNN的肌电图信号神经肌肉疾病分类
信号处理
2024-10-30 v3
摘要
肌萎缩侧索硬化症(ALS)与肌病在神经肌肉疾病诊断领域构成显著挑战。本研究采用先进深度学习技术以解决ALS与肌病这两种致残性疾病的检测问题。我们的方法始于从原始肌电图(EMG)信号中提取信息性特征，利用对数谱与差分对数谱，其捕捉信号的频率内容以及谱与时间特征。随后，我们应用一种深度学习模型SpectroEMG-Net，结合卷积神经网络(CNN)与注意力机制对三类进行分类。我们方法的鲁棒性得到严格评估，展现出在肌病、正常与ALS三类间区分的卓越性能，总体准确率达92%。本研究标志着通过数据驱动的多类分类方法应对神经肌肉疾病诊断挑战的贡献，为早期与准确检测的潜力提供宝贵见解。
引用
@article{arxiv.2309.10483,
title = {EMG Signal Classification for Neuromuscular Disorders with Attention-Enhanced CNN},
author = {Md. Toufiqur Rahman and Minhajur Rahman and Celia Shahnaz},
journal= {arXiv preprint arXiv:2309.10483},
year = {2024}
}
备注
We have identified an error in the methodology and calculations presented in our paper, which affects the validity of our results. We plan to correct these issues and resubmit once we have verified the findings