EKG sinyallerinden konjestif kalp yetmezliği ve aritmi teşhisi için yeni bir yaklaşım: İ-1B-YİÖ+LSTM
[ X ]
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Siirt Üniversitesi
Access Rights
info:eu-repo/semantics/openAccess
Abstract
Vücudun en hassas organlarından biri olan kalp, yaşamın sürdürülmesinde en hayati role sahip organlardandır. Elektrokardiyografi, kalp kaslarının kasılma ve gevşemesiyle oluşan elektriksel aktiviteyi ölçen ve kaydeden bir tanı yöntemidir. Kalpteki elektriksel aktivitelerin genliklerinin zaman düzleminde kaydedilmesiyle elektrokardiyogram (EKG) sinyalleri elde edilir. EKG sinyalleri, genellikle ritim bozukluğu gibi hastalıkların teşhisi için kullanılmaktadır. Bu çalışmada EKG sinyallerinden Konjestif Kalp Yetmezliği (KKY) ve Aritmi (ARR) teşhisi için, İndirgenmiş Bir Boyutlu Yerel İkili Örüntüler (İ-1B-YİÖ) ve Uzun Kısa Zamanlı Bellek Ağı (LSTM) tabanlı yeni bir yaklaşım önerilmiştir. EKG sinyalleri için İ-1B-YİÖ denilen dönüşüm yöntemi önerilmiştir. Bu yöntem tek boyutlu sinyallerin üzerindeki her bir değerin kendi komşuları ile oluşturdukları büyüklük-küçüklük bilgisini kullanır. Bu yöntem ile elde edilen indirgenmiş sinyal gruplarından hesaplanan 1B yerel ikili örüntülere ait histogramlar birleştirilerek LSTM modeline giriş olarak verilmektedir. Hem tek yönlü hem de çift yönlü İ-1B-YİÖ dönüşümü uygulanan EKG sinyallerine LSTM uygulanmıştır Önerilen yaklaşımı test etmek için; KKY, ARR ve Normal Sinüs Ritmi (NSR) olmak üzere 3 farklı durum bilgisi içeren toplam 972 EKG sinyali kullanılmıştır. Sinyaller MIT-BIH ve BIDMC veri bankalarından alınmıştır. Farklı senaryolarda denemeler gerçekleştirilmiştir. Elde edilen başarı oranları literatürdeki diğer çalışmalarla karşılaştırıldığında önerilen yaklaşım ile kabul edilebilir yüksek sonuçların elde edildiği görülmektedir. Elde edilen EKG tanı performans değerleri %96,80 ile %99,79 arasında değişim göstermiştir. Önerilen hastalık teşhis sisteminin, farklı medikal sinyallere de uygulanabileceği düşünülmektedir.
The heart, one of the body's most sensitive organs, is one of the organs that has the most vital role in maintaining life. Electrocardiography is a diagnostic technique that measures and records the electrical activity induced by heart muscle contraction and relaxation. Electrocardiogram (ECG) signals are obtained by recording the amplitudes of electrical activities in the heart in the time domain. ECG signals are generally used to diagnose diseases such as arrhythmias. In this study, a new approach based on Down Sampling-One Dimensional-Local Binary Pattern (1D-DS-LBP) and Long Short-Term Memory (LSTM) is proposed for the diagnosis of Congestive Heart Failure (CHF) and Arrhythmia (ARR) from ECG signals. For ECG signals, a conversion method called 1D-DS-LBP has been proposed. This method uses the larger or smaller information created by each value on one-dimensional signals with their neighbors. The histograms of the 1D local binary patterns calculated from the reduced signal groups obtained by this method are combined and given as an input to the LSTM model. LSTM was applied to the ECG signals that applied both unidirectional and bidirectional I-1D-YIS conversion. To test the proposed approach; A total of 972 ECG signals containing 3 different status information, CHF, ARR and Normal Sinus Rhythm (NSR), were used. Signals were taken from the MIT-BIH and BIDMC databases. Experiments were carried out in different scenarios. When the success rates obtained are compared with other studies in the literature, it is seen that acceptable high results are obtained with the proposed approach. The obtained ECG diagnostic performance values varied between %96,80 and %99,79. It is thought that the proposed disease diagnosis system can also be applied to different medical signals.
The heart, one of the body's most sensitive organs, is one of the organs that has the most vital role in maintaining life. Electrocardiography is a diagnostic technique that measures and records the electrical activity induced by heart muscle contraction and relaxation. Electrocardiogram (ECG) signals are obtained by recording the amplitudes of electrical activities in the heart in the time domain. ECG signals are generally used to diagnose diseases such as arrhythmias. In this study, a new approach based on Down Sampling-One Dimensional-Local Binary Pattern (1D-DS-LBP) and Long Short-Term Memory (LSTM) is proposed for the diagnosis of Congestive Heart Failure (CHF) and Arrhythmia (ARR) from ECG signals. For ECG signals, a conversion method called 1D-DS-LBP has been proposed. This method uses the larger or smaller information created by each value on one-dimensional signals with their neighbors. The histograms of the 1D local binary patterns calculated from the reduced signal groups obtained by this method are combined and given as an input to the LSTM model. LSTM was applied to the ECG signals that applied both unidirectional and bidirectional I-1D-YIS conversion. To test the proposed approach; A total of 972 ECG signals containing 3 different status information, CHF, ARR and Normal Sinus Rhythm (NSR), were used. Signals were taken from the MIT-BIH and BIDMC databases. Experiments were carried out in different scenarios. When the success rates obtained are compared with other studies in the literature, it is seen that acceptable high results are obtained with the proposed approach. The obtained ECG diagnostic performance values varied between %96,80 and %99,79. It is thought that the proposed disease diagnosis system can also be applied to different medical signals.
Description
Fen Bilimleri Enstitüsü, Elektrik-Elektronik Mühendisliği Ana Bilim Dalı
Keywords
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control, Elektrik ve Elektronik Mühendisliği
