Uyku evrelerinin sýnýflandýrýlmasýnda EEG ve EOG sinyallerinin karþýlaþtýrýlmasý

Yaþamýn en önemli parçasý olan uykunun deðeri, uykusuzluðun neden olduðu saðlýk sorunlarýnýn ortaya çýkmasýyla birlikte artmaktadýr. Bu sorunu çözmek için uyku evrelerinde ortaya çýkan farklý sinyal kalýplarýný yorumlamak son derece önemlidir. Bu amaca ulaþmak için uyku evrelerinin otomatik olarak puanlanmasýný saðlayan sistemler oluþturulur. Uyku puanlamasýnda uyuyan kiþinin elektrofizyolojik sinyalleri dikkate alýnarak uyku hakkýnda deðerli bilgiler elde edilir. Çalýþmada uyku alanýnda çalýþan araþtýrmacýlara açýk eriþim olarak sunulan ISRUC-Sleep veri seti kullanýlmýþtýr. Çalýþmanýn temel amacý, uyku evrelerinin sýnýflandýrýlmasýnda elektroensefalografi (EEG) ve elektrookülografi (EOG) biyosinyallerinin etkisini araþtýrmaktýr. Analiz, ISRUC platformuna ait üç farklý grubu tanýmlayan veri setinin üçüncü grubu dikkate alýnarak gerçekleþtirilmiþtir. Veri setindeki alt grup_3'ün 10 katýlýmcýsý dikkate alýnmýþtýr. Etkili öznitelikler çýkarýlarak ve farklý sýnýflandýrma yöntemleri uygulanarak aþamalarýn sýnýflandýrýlmasýnda EEG veya EOG sinyallerinden hangisinin daha iyi olduðu araþtýrýlmýþtýr. Kullanýlan sýnýflandýrma yöntemlerinin performans deðerlendirmesi açýsýndan önceki çalýþmamýzda sunulan yeni Roza metriði uygulanmýþtýr. Welch öznitelik çýkarma yöntemi ve toplu aðaç sýnýflandýrma tekniði sayesinde uyku evrelerinin sýnýflandýrýlmasýnda EEG sinyallerinin EOG'dan daha baþarýlý olduðu kanýtlanmýþtýr. Bu uyku evreleri EEG sinyallerini kullanarak %77,7 baþarý oranýyla sýnýflandýrýlmýþtýr.

Comparison of EEG and EOG signals in classification of sleep stages

The value of sleep, which is the most significant part of life, increases with the emergence of health problems caused by insomnia. To solve this problem, it is extremely important to interpret the different signal patterns that occur during sleep stages. In order to achieve this goal, systems are created that provide automatic scoring of sleep stages. In sleep scoring, valuable information about sleep is obtained by considering the electrophysiological signals of the sleeper. The ISRUC-Sleep dataset, which was presented as open access to researchers working in the field of sleep, was used in the study. The main goal of the study is to investigate the effect of electroencephalography (EEG) and electrooculography (EOG) biosignals in the classification of sleep stages. The analysis was carried out by considering the third group of the data set, which defines three different groups belonging to the ISRUC platform. The 10 participants of subgrup_3 in the dataset were considered. By extracting effective features and applying different classification methods, it was investigated which one of the EEG or EOG signals was better in the classification of stages. In terms of performance evaluation of the classification methods used, the new Roza metric presented in our previous study was applied. It has been proven that EEG signals are more successful than EOG in the classification of sleep stages, thanks to the Welch feature extraction method and the ensemble of bagged tree classification technique. These sleep stages were classified by using EEG signals with a success rate of 77.7%.