Elias Ebrahimzadeh1, Alireza Foroutan2, Mohammad Shams3, Raheleh Baradaran4, Lila Rajabion5, Mohammadamin Joulani6, Farahnaz Fayaz7. 1. School of Electrical and Computer Engineering, College of Engineering, University of Tehran, Tehran, Iran; Seaman Family MR Research Centre, University of Calgary, Calgary, Alberta, Canada; Department of Clinical Neurosciences, University of Calgary, Calgary, Alberta, Canada. Electronic address: e_ebrahimzadeh@ut.ac.ir. 2. Department of Biomedical Engineering, Faculty of Engineering, University of Isfahan, Isfahan, Iran. 3. Department of Biomedical Engineering, George Washington University, Washington D.C., USA. 4. Department of Anatomy and Cell Biology, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran. 5. Department of IT, College of Business, University of South Florida Sarasota-Manatee, USA. 6. Student Research Committee, Faculty of Medicine, Iran University of Medical Sciences, Tehran, Iran. 7. Biomedical Engineering Department, School of Electrical Engineering, Payame Noor University of North Tehran, Tehran, Iran.
Abstract
BACKGROUND AND OBJECTIVE: Taking into consideration the critical importance of Sudden cardiac death (SCD), as it could be the first and the last heart condition to be diagnosed in a person while continuing to claim millions of lives around the world, prediction of sudden cardiac death has increasingly been regarded as a matter of substantive significance. This study does not seek to once again define features to predict and detect SCD, as there already has been adequate discussion addressing feature extraction in our previous works and other recent studies. What seems to be lacking attention is the need for an appropriate strategy to manage the extracted features to such an extent that the best classification is presented. To this end, deploying a suitable tactic to select extracted features could bring about outstanding results compared to other works in the literature. METHODS: This research has accordingly applied a novel and automated approach to Local Feature Subset Selection through the most rigorous methodologies, which have formerly been developed in previous works of this team, for extracting features from nonlinear, time-frequency and classical processes. We are therefore enabled to select features that differ from one another in each minute before the incident through the agency of optimal feature selection in each one-minute interval of the signal. RESULTS: Using the proposed algorithm, SCD can be predicted 13 min before the onset, thus, better results are achieved compared to other techniques proposed in this field. Additionally, through defining a Utility Function and employing statistical analysis the alarm threshold has been effectively determined as 84% for the prediction accuracy. Having selected the best combination of features, based on their ability to generate the highest classification accuracy, the two classes are classified by means of the Multilayer Perceptron (MLP), K- Nearest Neighbor (KNN) the Support Vector Machine (SVM), and the Mixture of Expert (ME) classifier. The Mixture of Experts classification yielded more precise class discrimination than other well-known classifiers. The performance of the proposed method was evaluated using the MIT-BIH database which led to sensitivity, specificity, and accuracy of 84.24%, 85.71%, and 82.85%, respectively for thirteen one-minute. CONCLUSIONS: The outcome of the obtained prediction would be analyzed and compared to other results. The most applicable and effective features would subsequently be presented according to the number of times they have been chosen. Finally, principal features in each time interval are discussed and the importance of each type of processing will be drawn into focus. The results indicate the significant capacity of the proposed method for predicting SCD from Electrocardoigram (ECG) signals as well as selecting the appropriate processing method at any time before the incident.
BACKGROUND AND OBJECTIVE: Taking into consideration the critical importance of Sudden cardiac death (SCD), as it could be the first and the last heart condition to be diagnosed in a person while continuing to claim millions of lives around the world, prediction of sudden cardiac death has increasingly been regarded as a matter of substantive significance. This study does not seek to once again define features to predict and detect SCD, as there already has been adequate discussion addressing feature extraction in our previous works and other recent studies. What seems to be lacking attention is the need for an appropriate strategy to manage the extracted features to such an extent that the best classification is presented. To this end, deploying a suitable tactic to select extracted features could bring about outstanding results compared to other works in the literature. METHODS: This research has accordingly applied a novel and automated approach to Local Feature Subset Selection through the most rigorous methodologies, which have formerly been developed in previous works of this team, for extracting features from nonlinear, time-frequency and classical processes. We are therefore enabled to select features that differ from one another in each minute before the incident through the agency of optimal feature selection in each one-minute interval of the signal. RESULTS: Using the proposed algorithm, SCD can be predicted 13 min before the onset, thus, better results are achieved compared to other techniques proposed in this field. Additionally, through defining a Utility Function and employing statistical analysis the alarm threshold has been effectively determined as 84% for the prediction accuracy. Having selected the best combination of features, based on their ability to generate the highest classification accuracy, the two classes are classified by means of the Multilayer Perceptron (MLP), K- Nearest Neighbor (KNN) the Support Vector Machine (SVM), and the Mixture of Expert (ME) classifier. The Mixture of Experts classification yielded more precise class discrimination than other well-known classifiers. The performance of the proposed method was evaluated using the MIT-BIH database which led to sensitivity, specificity, and accuracy of 84.24%, 85.71%, and 82.85%, respectively for thirteen one-minute. CONCLUSIONS: The outcome of the obtained prediction would be analyzed and compared to other results. The most applicable and effective features would subsequently be presented according to the number of times they have been chosen. Finally, principal features in each time interval are discussed and the importance of each type of processing will be drawn into focus. The results indicate the significant capacity of the proposed method for predicting SCD from Electrocardoigram (ECG) signals as well as selecting the appropriate processing method at any time before the incident.