A new dataset for EEG abnormality detection MTOUH
Yükleniyor...
Tarih
2022
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
İrem TAŞCI
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
Elektroensefalogram (EEG), beyindeki elektriksel aktivitenin izlenmesi için yaygın olarak kullanılmaktadır. EEG sinyallerinin hekimler tarafından incelenmesi yorucu ve zaman alıcıdır. Bu nedenle, algılama doğruluğunu artırmak için makine öğrenme teknikleri kullanılabilir. Bu çalışmada 35 kanal, 10575x15 saniye normal ve 11240x15 saniye anormal EEG sinyalinden oluşan 2 sınıflı veri seti oluşturulmuştur. Bu very seti Turgut Özal Üniversitesi Malatya Eğitim Araştırma Hastanesi’ ne 2021 yılında başvuran hastaların EEG sinyalleri incelenerek elde edilmiştir. Çalışmada istatistiksel özellik çıkarımı tabanlı bir model önerilmiştir. Önerilen modele komşu bileşen analizi kullanılarak öznitelik vektörü indirgemesi yapıldıktan sonra destek vektör makineleri kullanılarak sınıflandırma yapılmıştır. 35 kanaldan en yüksek doğruluk P4O2 kanalında elde edilmiştir. P4O2 kanalı için doğruluk, duyarlılık, özgüllük, kesinlik ve f-skoru sırasıyla %81.3,%78.9, %83.7, %82.0 ve %80.4 olarak elde edilmiştir.
Electroencephalogram (EEG) is widely used for monitoring electrical activity in the brain. Analyzing EEG signals by physicians is tiring and time-consuming. Therefore, machine learning techniques can be used to improve detection accuracy. This study created a 2-class data set consisting of 35 channels, 10575x15 seconds of normal and 11240x15 seconds of abnormal EEG signals. This data set was obtained by examining the EEG signals of the patients who applied to Malatya Turgut Ozal University Malatya Research and Training Hospital in 2021. In the study, a statistical feature extraction-based model is proposed. After the feature vector reduction was performed using the neighboring component analysis to the proposed model, the classification was made using the support vector machines. The highest accuracy out of 35 channels was obtained in the P4O2 channel. Accuracy, sensitivity, specificity, precision and f-score for the P4O2 channel were 81.3%, 78.9%, 83.7%, 82.0% and 80.4%, respectively.
Electroencephalogram (EEG) is widely used for monitoring electrical activity in the brain. Analyzing EEG signals by physicians is tiring and time-consuming. Therefore, machine learning techniques can be used to improve detection accuracy. This study created a 2-class data set consisting of 35 channels, 10575x15 seconds of normal and 11240x15 seconds of abnormal EEG signals. This data set was obtained by examining the EEG signals of the patients who applied to Malatya Turgut Ozal University Malatya Research and Training Hospital in 2021. In the study, a statistical feature extraction-based model is proposed. After the feature vector reduction was performed using the neighboring component analysis to the proposed model, the classification was made using the support vector machines. The highest accuracy out of 35 channels was obtained in the P4O2 channel. Accuracy, sensitivity, specificity, precision and f-score for the P4O2 channel were 81.3%, 78.9%, 83.7%, 82.0% and 80.4%, respectively.
Açıklama
Anahtar Kelimeler
Electroencephalography signals classification,, SVM,, Machine Learning, Elektroensefalografi sinyal sınıflandırması, DVM, Makine Öğrenmesi
Kaynak
Turkish Journal of Science and Technology
WoS Q Değeri
Scopus Q Değeri
Cilt
17
Sayı
1
Künye
TAŞCI, İ., TASCİ, B., DOGAN, S., & TUNCER, T. (2022). A new dataset for EEG abnormality detection MTOUH. Turkish Journal of Science and Technology, 17(1), 135-141.
