本論文提出一人工智慧物聯網心電圖分析系統，具備完整之物聯網架構，達成即時心電訊號辨識及監控系統，包含心律偵測器、智慧型裝置使用者介面、雲端人工智慧分析網路，並將其中之雲端人工智慧分析網路以硬體實現其一維卷積神經網路架構，達成邊緣運算、硬體加速之目的。且心律偵測器、智慧型裝置使用者介面皆依照衛生福利部食品藥物管理署(TFDA)上市前臨床測試之前置醫療電器設備基本安全標準之相關設計，進行可用性評估報告、功能性試驗(IEC60601-2-47)、電磁相容性試驗及風險管理。
透過人工智慧物聯網心電圖分析系統，不僅可以即時的監控使用者之心電圖訊號(Electrocardiogram, ECG)更可以直接獲取經雲端人工智慧神經網路分析之結果，因心血管疾病可由心電圖進行辨識，此一維卷積神經網路主要分析心房早期收縮(APC,Atrial Premature Contraction)及心室早期收縮(VPC,Ventricular Premature Contraction)此兩項心血管疾病，能即時且快速的將異常波形抓出提供給醫師，有助於輔助醫師進行心血管疾病上之辨識。
本論文之系統架構為符合醫療安規標準設計之心律偵測器裝置，內含前端心電圖訊號感測電路、具藍芽5.0之微控制器，搭配智慧型裝置之使用者介面以及雲端建置之一維卷積神經網路分析以及其神經網路硬體化之晶片，先利用前端感測電路進行醫療等級的訊號感測，透過具藍芽5.0之微控制器，可依據使用者情境，分別在兩種模式下運行，離線模式將訊號記錄在記憶卡中，即時模式則是將心電圖訊號透過藍芽5.0傳送至智慧型裝置之使用者介面，進行心電圖訊號之即時監控、辨識。
雲端建置之一維卷積神經網路分析系統，其架構設計為因應硬體化，層數、參數、暫存值皆經特殊設計，且已實際硬體化，此一維卷積神經網路模型實際運用於臨床測試，神經網路模型訓練之資料庫來源包含MIT-BIH以及衛生福利部台南醫院進行人體試驗收案之病人，在總數9萬多筆的資料中，正確率(Accuracy)為99.49% ，心室早期收縮的準確率(Precision)為97.31%，心房早期收縮的準確率為95.5%。

This paper presents an electrocardiogram (ECG) analysis system with artificial intelligence of things (AIoT) design. The system is composed of ECG patch, the smart phone APP and Cloud artificial intelligence network. ECG patch and APP are implemented according to Guidance for Pre-Clinical Testing of Electrocardiograph [2] which was announced by Taiwan Food and Drug Administration (TFDA) including usability testing, performance test, electromagnetic compatibility test, and risk management.
The users can monitor the real-time ECG signal on the smart phone APP. Also the ECG analysis of artificial intelligence is based on one-dimensional convolution neural network (CNN) on cloud server with the ability to classify three kinds of ECG, including normal beat, atrial premature complex (APC) and ventricular premature contraction (VPC). The cloud server includes signal preprocessing, feature extraction, and CNN model, and return the result to the APP immediately.
Considering of edge computing, hardware of the CNN model on cloud server is implemented by the chip fabricated with 0.18um CMOS technology. Except the CNN model, the Universal Asynchronous Receiver/Transmitter (UART) is also implemented. The chip which based on UART can be embedded in the whole system for the purpose of less cloud power consumption and the hardware acceleration. The proposed system is verified by clinical trial. And the CNN model is tested on two different database including MIT database and patients in Ministry of Health and Welfare Tainan Hospital. The accuracy of the proposed system is about 99.49% by the verification on the software. Simultaneously, and the precision of VPC is 97.31% and the precision of APC is 95.5%

[1] 108年中華民國國人死因統計 台灣衛生福利部統計處
[2] 心電圖描記器臨床前測試基準 [Online] 衛生福利部食品藥物管理署
Available:https://info.fda.gov.tw/tc/siteListContent.aspx?sid=3787&id=30858
[3] 醫療器材採認標準資料庫 [Online].
Available:https://mdlicense.itri.org.tw/MDDB/(S(mq40x345qf0fn1j3jcpbsu55))/Recognized/Recognized.aspx
[4] IEC60601-2-47之解決方案 鯨揚科技
[5] SEGC4.0 [Online].
Available:https://www.whaleteq.com/tw/Products/Detail/18/SECG%204.0#tag=proOverview
[6] CMRR3.0 [Online].
Available:https://www.whaleteq.com/tw/Products/Detail/40/CMRR%203.0+#tag=proOverview
[7] ISO14971:2019 Medical devices — Application of risk management to medical devices
[8] Raytac [Online]
Available: https://www.raytac.com/
[9] J. Pan and W. J. Tompkins, "A real-time QRS detection algorithm", IEEE Trans. Biomed. Eng., vol. BME-32, pp. 230-236, 1985
[10] Hannun, A.Y., Rajpurkar, P., Haghpanahi, M. et al. Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network. Nat Med 25, 65–69 (2019)
[11] MIT-BIH [Online]
Available : https://www.physionet.org/physiobank/database/mitdb/
[12] D. Li, J. Zhang, Q. Zhang and X. Wei, "Classification of ECG signals based on 1D convolution neural network," 2017 IEEE 19th International Conference on e-Health Networking, Applications and Services (Healthcom), Dalian, pp. 1-6,2017
[13] Diederik P. Kingma, Jimmy Ba “Adam: A Method for Stochastic Optimization " 3rd International Conference for Learning Representations, San Diego, pp1-15,2015
[14] Y. Lin, C. Chuang, C. Yen, S. Huang, J. Chen and S. Lee, "An AIoT Wearable ECG Patch with Decision Tree for Arrhythmia Analysis," 2019 IEEE Biomedical Circuits and Systems Conference (BioCAS), Nara, Japan, pp. 1-4,2019
[15] U. R. Acharya et al., "Automated characterization of arrhythmias using nonlinear features from tachycardia ECG beats," 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Budapest, pp1-4, 2016
[16] Ullah, A.; Anwar, S.M.; Bilal, M.; Mehmood, R.M. Classification of Arrhythmia by Using Deep Learning with 2-D ECG Spectral Image Representation. Remote Sens. pp1-4, 2020, 12
[17] D. SS and D. J. Auxillia, "Classification of ECG using convolutional neural network (CNN)," 2019 International Conference on Recent Advances in Energy-efficient Computing and Communication (ICRAECC), Nagercoil, India, pp. 1-6, 2019
[18] Y. He, X. Zhang and J. Sun, "Channel Pruning for Accelerating Very Deep Neural Networks," 2017 IEEE International Conference on Computer Vision (ICCV), Venice, pp. 1398-1406,2017,
[19] S. Kiranyaz, T. Ince and M. Gabbouj, "Real-Time Patient-Specific ECG Classification by 1-D Convolutional Neural Networks," in IEEE Transactions on Biomedical Engineering, vol. 63, no. 3, pp1-12 5, March 2016
[20] M. Zubair, J. Kim and C. Yoon, "An Automated ECG Beat Classification System Using Convolutional Neural Networks," 2016 6th International Conference on IT Convergence and Security (ICITCS), Prague, pp. 1-5,2016
[21] S. Kiranyaz, T. Ince, O. Abdeljaber, O. Avci and M. Gabbouj, "1-D Convolutional Neural Networks for Signal Processing Applications," ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Brighton, United Kingdom, pp. 8360-8364,2019,
[22] J. Huang et al., "Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, pp. 3296-3297,2017
[23] Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998.
[24] Diederik P. Kingma, Jimmy Ba “Adam: A Method for Stochastic Optimization " 3rd International Conference for Learning Representations, San Diego, pp1-15,2015
[25] Yan Fang, Chao Li, Lijun Sun, Design of an Early Warning System for Patients with Cardiovascular Diseases Under Mobile Environment, pp1-4, 2016.08.258. pp
[26] K. M. C. Tan and R. A. Maxion, ""Why 6?" Defining the operational limits of stide, an anomaly-based intrusion detector," Proceedings 2002 IEEE Symposium on Security and Privacy, Berkeley, CA, USA, pp. 188-201, 2002
[27] Bing Xu, Naiyan Wang, Tianqi Chen, Mu Li “Empirical Evaluation of Rectified Activations in Convolutional Network” arxiv, pp1-5 2015
[28] Varun Ranganathan, S. Natarajan “A New Backpropagation Algorithm without arxiv, pp1-15 2018
[29] M. Z. U. Rahman, R. A. Shaik and D. V. R. K. Reddy, "An Efficient Noise Cancellation Technique to Remove Noise from the ECG Signal Using Normalized Signed Regressor LMS Algorithm," 2009 IEEE International Conference on Bioinformatics and Biomedicine, Washington, DC, pp. 257-260, 2009,
[30] S. Hansun, "A new approach of moving average method in time series analysis," 2013 Conference on New Media Studies (CoNMedia), Tangerang, 2013, pp. 1-4