Sudha G, Saravanan N, Muthalakshmi M, Birunda M
Department of Biomedical Engineering, Muthayammal Engineering College, Rasipuram, Namakkal, Tamil Nadu India.
Department of Biotechnology, Muthayammal Engineering College, Rasipuram, Namakkal, Tamil Nadu India.
Health Inf Sci Syst. 2024 Mar 15;12(1):25. doi: 10.1007/s13755-024-00284-9. eCollection 2024 Dec.
Alzheimer's disease is an incurable neurological disorder that damages cognitive abilities, but early identification reduces the symptoms significantly. The absence of competent healthcare professionals has made automatic identification of Alzheimer's disease more crucial since it lessens the amount of work for staff members and improves diagnostic outcomes. The major aim of this work is "to develop a computer diagnostic scheme that makes it possible to identify AD using the Electroencephalogram (EEG) signal". Therefore, Dynamically Stabilized Recurrent Neural Network Optimized with Artificial Gorilla Troops espoused Alzheimer's Disorder Detection using EEG signals (DSRNN-AGTO-ADD) is proposed in this paper. Here, Dynamic Context-Sensitive Filter (DCSF) is considered to eliminate the noise, and interference from the EEG signal. Then Adaptive and Concise Empirical Wavelet Transform (ACEWT) is utilized to separate the filtered signals from the frequency bands, and to feature extraction from the EEG signals. Signal's characteristics, like logarithmic bandwidth power, standard deviation, variance, kurtosis, mean energy, mean square, norm are combined to ACEWT method to create feature vectors and enhance diagnostic performance. After that, the extracted features are fed to Dynamically Stabilized Recurrent Neural Network (DSRNN) for task classification. Weight parameter of DSRNN is enhanced using Artificial Gorilla Troops Optimization Algorithm (AGTOA). The proposed DSRNN-AGTOA-ADD algorithm is activated in MATLAB. The metrics including accuracy, specificity, sensitivity, precision, computation time, ROC are examined for AD diagnosis. The performance of the proposed DSRNN-AGTOA-ADD approach attains 12.98%, 5.98% and 23.45% high specificity; 29.98%, 23.32% and 19.76% lower computation Time and 29.29%, 8.365%, 8.551% and 7.915% higher ROC compared with the existing methods.
阿尔茨海默病是一种无法治愈的神经紊乱疾病，会损害认知能力，但早期识别可显著减轻症状。由于缺乏专业的医疗保健人员，阿尔茨海默病的自动识别变得更加关键，因为它减少了工作人员的工作量并改善了诊断结果。这项工作的主要目标是“开发一种计算机诊断方案，能够使用脑电图（EEG）信号识别阿尔茨海默病”。因此，本文提出了一种基于人工大猩猩群优化的动态稳定递归神经网络用于基于EEG信号的阿尔茨海默病检测（DSRNN - AGTO - ADD）。这里，采用动态上下文敏感滤波器（DCSF）来消除EEG信号中的噪声和干扰。然后，利用自适应简洁经验小波变换（ACEWT）将滤波后的信号从频带中分离出来，并从EEG信号中进行特征提取。信号的特征，如对数带宽功率、标准差、方差、峰度、平均能量、均方、范数等，与ACEWT方法相结合以创建特征向量并提高诊断性能。之后，将提取的特征输入到动态稳定递归神经网络（DSRNN）进行任务分类。使用人工大猩猩群优化算法（AGTOA）增强DSRNN的权重参数。所提出的DSRNN - AGTOA - ADD算法在MATLAB中实现。对包括准确率、特异性、灵敏度、精度、计算时间、ROC等指标进行阿尔茨海默病诊断检查。与现有方法相比，所提出的DSRNN - AGTOA - ADD方法的性能实现了高12.98%、5.98%和23.45%的特异性；低29.98%、23.32%和19.76%的计算时间以及高29.29%、8.365%、8.551%和7.915%的ROC。
