Scholarlink

<p>Migraine is a&#xa0;common and chronic neurological disorder, and its diagnosis is based on the criteria of the International Classification of Headache Disorders (ICHD). However, delays in the accurate diagnosis of migraine may occur due to similar symptoms; this can negatively impact both the patient’s quality of life and healthcare economics. The aim of this study is to develop an approach that contributes to artificial intelligence-based migraine diagnostic systems by analyzing electroencephalography (EEG) signals obtained from migraine patients. In the proposed study, two distinct datasets were used, consisting of EEG signals from migraine patients and a&#xa0;healthy control group with similar characteristics. The EEG signals were analyzed in their raw form and were decomposed into sub-bands using adaptive signal processing methods such as Robust Local Mean Decomposition (RLMD), Circular Single-Spectrum Analysis (CiSSA), and Variational Mode Decomposition (VMD). The time-domain, frequency-domain, nonlinear, and statistical distribution features obtained from the analysis that positively influence classification performance were identified using the Least Absolute Shrinkage and Selection Operator (LASSO) feature selection method. Using the identified features, an artificial neural network (ANN), a&#xa0;support vector machine (SVM), and a&#xa0;random forest (RF) were trained to classify migraine patients and the healthy control group. According to the analysis results, the band power feature had a&#xa0;significant effect in distinguishing between the migraine and healthy groups in the C4 and F7 channels. The RLMD method yielded the most successful results in classification using ANN, SVM, and RF, with an mean accuracy rate of 94.58% in both datasets. This rate was 87.40% for the CiSSA method and 89.58% for the VMD method. The mean accuracy rate for raw EEG signals not decomposed into sub-bands was 83.64%. The application of adaptive signal processing methods resulted in an approximate 11% increase in accuracy rates. The results obtained in the proposed study demonstrate that adaptive signal processing techniques using EEG signals and machine learning algorithms are effective in migraine analysis. This approach enables the development of decision support systems that complement the ICHD criteria, thereby facilitating clinicians’ ability to reach an accurate diagnosis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EEG-Based migraine classification with machine learning algorithms: comparative analysis of adaptive decomposition techniques

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • Fırat Orhanbulucu,
  • Fatma Latifoğlu

摘要

Migraine is a common and chronic neurological disorder, and its diagnosis is based on the criteria of the International Classification of Headache Disorders (ICHD). However, delays in the accurate diagnosis of migraine may occur due to similar symptoms; this can negatively impact both the patient’s quality of life and healthcare economics. The aim of this study is to develop an approach that contributes to artificial intelligence-based migraine diagnostic systems by analyzing electroencephalography (EEG) signals obtained from migraine patients. In the proposed study, two distinct datasets were used, consisting of EEG signals from migraine patients and a healthy control group with similar characteristics. The EEG signals were analyzed in their raw form and were decomposed into sub-bands using adaptive signal processing methods such as Robust Local Mean Decomposition (RLMD), Circular Single-Spectrum Analysis (CiSSA), and Variational Mode Decomposition (VMD). The time-domain, frequency-domain, nonlinear, and statistical distribution features obtained from the analysis that positively influence classification performance were identified using the Least Absolute Shrinkage and Selection Operator (LASSO) feature selection method. Using the identified features, an artificial neural network (ANN), a support vector machine (SVM), and a random forest (RF) were trained to classify migraine patients and the healthy control group. According to the analysis results, the band power feature had a significant effect in distinguishing between the migraine and healthy groups in the C4 and F7 channels. The RLMD method yielded the most successful results in classification using ANN, SVM, and RF, with an mean accuracy rate of 94.58% in both datasets. This rate was 87.40% for the CiSSA method and 89.58% for the VMD method. The mean accuracy rate for raw EEG signals not decomposed into sub-bands was 83.64%. The application of adaptive signal processing methods resulted in an approximate 11% increase in accuracy rates. The results obtained in the proposed study demonstrate that adaptive signal processing techniques using EEG signals and machine learning algorithms are effective in migraine analysis. This approach enables the development of decision support systems that complement the ICHD criteria, thereby facilitating clinicians’ ability to reach an accurate diagnosis.