딥러닝을 이용한 심전도 분류 알고리즘에 관한 연구
A Study of ECG Classification Algorithm Using Deep Learning
0
조회
0
다운로드
3
인용
0
참고문헌
키워드
초록
심전도는 심장 활동에 대한 전기신호의 위치 수치의 변화를 시간대 별로 저장하는 방법으로 심장의 정상 또는 이상상태를 분석하는 검사 방법이다. 기존의 심전도 분석 알고리즘에는 주파수 대역 통과 필터를 이용한 규칙기반의 분석이 사용되었으나 최근 인공지능 기술을 적용한 딥러닝, 기계학습 같은 알고리즘들이 연구되고 있다. 본 논문에서는 심전도 data set에 고속 푸리에 변환을 전처리로 수행하고 0.67Hz-50Hz의 대역필터를 적용하고 4초간의 심전도 비트를 기준점의 이전 3초, 이후 1초간의 데이터로 재구성한다. 제안모델은 딥러닝의 합성곱 신경망 연산을 적용하고 overfitting 해결방안으로 dropout을 반복하는 모델을 제안하였다. 학습데이터는 MIT-BIH의 48개의 부정맥 data set 중 4개의 페이스메이커 제외한 44개 data set 학습하였다. 제안한 합성곱 신경망 모델의 연구결과는 평균 정확도 99.8%이며 평가데이터의 평균 정확도는 99.6%로 기존 연구대비 동등하거나 유사한 성능을 보였다. 본 논문에서 제안한 모델은 심장비트에 대한 분류이다. 그러나 학습모델을 개선하면 심방세동와 심실빈맥, 심실세동 등의 심전도 리듬 분류 연구에 도움이 될 것이다.
Abstract
The electrocardiogram(ECG) is a algorithm of recording the change in potential value for the activity of the heart by time and has been used as a testbed algorithm to analyze the normal/abnormal state of the heart. The ECG analysis algorithm uses a frequency band filter but recently deep learning and machine learning using artificial intelligence technology. In this paper, we used of ECG data, a band filter between 0.67Hz and 50Hz was used by fast Fourier Transform and the ECG data for four seconds were reconstructed into data for three seconds before and one second after the beat to be classified as one. As the learning algorithm of artificial intelligence, deep learning, convolution neural network operation was used. To prevent overfitting of the learning model, we developed a model that classifies ECG by repeating dropout. for training data, 70% were trained using 44 data set excluding pacemaker among 48 record set of MIT-BIH arrhythmia data set. As a result of the testbed, the average accuracy of the convolution neural network model was 99.8%, whereas the average accuracy of all classification was 99.6% as a result of the evaluation data, indication the it showed similar compared to the performance of the conventional ECG classification. In this paper, it is a classification for heart beat. but it is possible to contribute to the study on classification of atrial fibrillation as classification of ECG rhythms by learning models.
참고문헌 (17)
- S. Kiranyaz, "Real time patient specific ECG classification by 1-D convolutional neural networks", IEEE Transactions on Bio-medical Engineering, 61(3), pp.664~675 (2015)
- M. Zubair, "An automated ECG beat classification system using convolution neural networks", International Conference on IT Convergence and Security, pp.11~17 (2016)
- C. Tin, "Automated ECG classification using dual heartbeat coupling based on convolutional neural network", IEEE Transactions on Systems, Access, 6, pp.2746~2752 (2018)
- O. T. Inan, "Robust neural network based classification of PVCs using wavelet transform and timing interval features", IEEE Transactions on Bio-medical Engineering, 53(12), pp.2507~2515 (2006)
- R. Daniele, "Deep learning for health informatics", IEEE Journal of Bio-medical and Health Informatics, 21(1), pp.4~21 (2017)
- R. J. Martis, "Representation learning : a review and new perspectives", IEEE Transactions on Bio-medical Engineering, 60(8), pp.1798~1828 (2013)
- N. Srivastava, "Dropout :a simple way to prevent neural networks from overfitting", Article in Journal of Machine Learning Research, 15(1), pp.1929~1958 (2014)
- "MIT-BIH Arrhythmia Database"[URL]
- S. Kwon, "Detection and classification of arrhythmia beat using convolution neural network", 석사, pp.2019
- K. Minami, "Real-time discrimination of ventricular tachyarrhthmia with fourier-transform neural network", IEEE Transactions on Bio-medical Engineering, 46(2), pp.179~185 (1999)
- D. Scolari, "Comparative study between DD-HMM and RBF in ventricular tachycardia and ventricular fibrillation recognition", IEEE Transactions on Bio-medical Engineering, 55(8), pp.213~217 (2008)
- L. Liu, "A deep-learning approach to ECG classification based on adversarial domain adaptation", pp.882~885 (2021), Computer algorithms Programs Bio-medical Engineering
- H. Wang, "An improved convolutional neural network based approach for automated heartbeat classification", Journal of Medical Systems, pp.1844~1856 (2019)
- S. Mustafa, "Design and implementation of an ultra-low energy FFT ASIC for processing ECG in cardiac pacemakers", IEEE Transactions on Bio-medical Engineering, 66(7), pp.983~987 (2019)
- Q. Wu, "ECG signal classification with binarized convolutional neural network", pp.1016~1025 (2020), Computer and Bio-medical Engineering
- 이세열, "방향성 가중치가 적용된 모듈기반의 개선된 하이브리드 침입시도탐지모델", 한국지식정보기술학회 논문지, 9(6), pp.685~690 (2014), 한국지식정보기술학회
- 이세열, "퍼지인식도 기반의 분산협력 침입시도탐지모델에 관한 연구", 한국지식정보기술학회 논문지, 16(1), pp.223~230 (2021), 한국지식정보기술학회[DOI]