Yuan Zhang1, Sen Liu2, Zhihui He3, Yuwei Zhang4, Changming Wang5. 1. Chongqing Key Laboratory of Nonlinear Circuits and Intelligent Information Processing, College of Electronic and Information Engineering, Southwest University, Chongqing, 400715, China. yuanzhang@swu.edu.cn. 2. Department of Oncology, Central Hospital Affiliated to Shandong First Medical University, Jinan, 250013, China. 3. Department of Pediatric Respiration, Chongqing Ninth People's Hospital, Chongqing, 400700, China. 4. School of Biological Science and Medical Engineering, Southeast University, Nanjing, 210096, China. zhangyuwei@seu.edu.cn. 5. Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing, 100053, China. superwcm@163.com.
Abstract
PURPOSE: Wearable devices in the scenario of connected home healthcare integrated with artificial intelligence have been an effective alternative to the conventional medical devices. Despite various benefits of wearable electrocardiogram (ECG) device, several deficiencies remain unsolved such as noise problem caused by user mobility. Therefore, an insensitive and robust classification model for cardiac arrhythmias detection system needs to be devised. METHODS: A one-dimensional seven-layer convolutional neural network (CNN) classification model with dedicated design of structure and parameters is developed to perform automatic feature extraction and classification based on large volume of original noisy signals. Record-based ten-fold cross validation scheme is devised for evaluation to ensure the independence of the training set and test set, and further improve the robustness of our method. RESULTS: The model can effectively detect cardiac arrhythmias, and can reduce the computational workload to a certain extent. Our experimental results outperform most recent literature on the cardiac arrhythmias classification with diagnostic accuracy of 0.9874, sensitivity of 0.9811, and specificity of 0.9905 for original signals; diagnostic accuracy of 0.9876, sensitivity of 0.9813, and specificity of 0.9907 for de-noised signals, respectively. CONCLUSION: The evaluation indicates that our proposed approach, which performs well on both original signals and de-noised signals, fits well with wearable ECG monitoring and applications.
PURPOSE: Wearable devices in the scenario of connected home healthcare integrated with artificial intelligence have been an effective alternative to the conventional medical devices. Despite various benefits of wearable electrocardiogram (ECG) device, several deficiencies remain unsolved such as noise problem caused by user mobility. Therefore, an insensitive and robust classification model for cardiac arrhythmias detection system needs to be devised. METHODS: A one-dimensional seven-layer convolutional neural network (CNN) classification model with dedicated design of structure and parameters is developed to perform automatic feature extraction and classification based on large volume of original noisy signals. Record-based ten-fold cross validation scheme is devised for evaluation to ensure the independence of the training set and test set, and further improve the robustness of our method. RESULTS: The model can effectively detect cardiac arrhythmias, and can reduce the computational workload to a certain extent. Our experimental results outperform most recent literature on the cardiac arrhythmias classification with diagnostic accuracy of 0.9874, sensitivity of 0.9811, and specificity of 0.9905 for original signals; diagnostic accuracy of 0.9876, sensitivity of 0.9813, and specificity of 0.9907 for de-noised signals, respectively. CONCLUSION: The evaluation indicates that our proposed approach, which performs well on both original signals and de-noised signals, fits well with wearable ECG monitoring and applications.
Authors: Johannes B Reitsma; Afina S Glas; Anne W S Rutjes; Rob J P M Scholten; Patrick M Bossuyt; Aeilko H Zwinderman Journal: J Clin Epidemiol Date: 2005-10 Impact factor: 6.437
Authors: Jen Hong Tan; Yuki Hagiwara; Winnie Pang; Ivy Lim; Shu Lih Oh; Muhammad Adam; Ru San Tan; Ming Chen; U Rajendra Acharya Journal: Comput Biol Med Date: 2018-01-02 Impact factor: 4.589
Authors: U Rajendra Acharya; Shu Lih Oh; Yuki Hagiwara; Jen Hong Tan; Muhammad Adam; Arkadiusz Gertych; Ru San Tan Journal: Comput Biol Med Date: 2017-08-24 Impact factor: 4.589