-
摘要: 連續血糖監測在糖尿病管理中具有重要的意義。目前糖尿病患者主要通過指尖采血或植入式微創傳感器監測血糖，但上述方法存在疼痛、成本昂貴、易感染等問題，因此，無創監測是實現連續血糖監測的理想技術。本文利用心電(ECG)信號，提出了一種血糖水平無創監測的方法：通過獲取12名志愿者共60 d 756160個ECG周期信號，利用遞歸濾波器實現ECG信號的濾波，并采用卷積神經網絡和長短期記憶網絡相結合(CNN-LSTM)的方法，實現了血糖水平的十分類監測，并通過實驗探索了個體建模和群體建模2種建模方式的差異。結果表明，在個體建模和群體建模的條件下，血糖監測精確率分別約達到80%和88%。其中群體建模10分類的F1值可達到0.95、0.88、0.91、0.85、0.92、0.88、0.86、0.86、0.87和0.86。研究表明，本文提出的基于ECG的無創血糖監測方法為實現血糖水平的實時、精準監測提供了一種有力的理論支撐與技術指導。Abstract: Continuous glucose monitoring is important in the management of diabetes. According to statistics, diabetes is the third chronic non-infectious disease that seriously endangers people's health, followed by tumor as well as cardiovascular and cerebrovascular diseases. In 2019, globally, there were a total of 460 million diabetics aged 20–79 years, which accounted for 9.1% of the total population in this cohort. Each figure is projected to increase to 592 million and by 10.1% respectively by 2035. Currently, the methods of blood glucose monitoring can be divided into invasive, minimally invasive, and noninvasive. The main methods for blood glucose monitoring include irregular sampling of fingertip blood or consecutive measurement of interstitial fluid glucose based on implantable sensors. However, these methods have some limitations, which include pain sensation, high cost, short service life, and susceptibility. Patients need to measure their blood glucose frequently. Invasive and minimally invasive monitoring will cause physical and psychological pain. Therefore, noninvasive monitoring is one of the most promising techniques for continuous monitoring of blood glucose, and it has a broad market prospect. In this study, the electrocardiogram (ECG signals) were used to achieve the noninvasive monitoring of blood glucose levels. First, 756160 ECG periodic signals of 12 volunteers for 60 d were obtained from the experiment. Second, the ECG signals were preprocessed using an infinite impulse response filter. Furthermore, a method combining convolutional neural networks and long short-term memory networks (CNN-LSTM) was proposed for blood glucose monitoring. In Addition, two modeling methods (individual modeling and group modeling) were investigated in this study. The results show that the precision of blood glucose monitoring under the condition of individual and group modeling is 80% and 88%, respectively. The F1-score of the group modeling can reach 0.95, 0.88, 0.91, 0.85, 0.92, 0.88, 0.86, 0.86, 0.87, and 0.86. Therefore, this study indicates that the proposed method based on ECG signals can provide powerful theoretical support and technical guidance for real-time and accurate blood glucose monitoring.
-
圖 3 ECG信號濾波前后圖像。（a）未濾波的ECG信號；（b）IIR濾波器去噪后的ECG信號
Figure 3. Images of ECG signals before and after filtering: (a) unfiltered ECG signal; (b) ECG signal followed by IIR filter
圖 4 不同志愿者在相同血糖水平下的一個ECG信號周期波形示例。（a）BG=5.9 mmol?L?1；（b）BG=8.1 mmol?L?1；（c）BG=10.5 mmol?L?1
Figure 4. ECG signal cycle waveforms at the same BG level for different subjects: (a) BG = 5.9 mmol?L?1; (b) BG = 8.1 mmol?L?1; (c) BG = 10.5 mmol?L?1
表 1 12名志愿者信息分布（人數）
Table 1. Quantity of volunteers with different body information
Gender Age bracket BMI Male Female ≤24 (24,40) ≥40 <18.5 (Low weight) [18.5,23) (Normal weight) ≥23 (Overweight) 5 7 4 5 3 3 6 3 
下載: 導出CSV
表 2 群體建模血糖分類標簽及數據量
Table 2. Blood glucose classification labels and data volumes upon group modeling
Blood glucose classification/(mmol?L?1) Labels Data Size Ration/% ≤5.6 0 70164 10.1 >5.6 and ≤6.2 1 75424 10.9 >6.2 and ≤6.6 2 66765 9.6 >6.6 and ≤7.2 3 75247 10.8 >7.2 and ≤7.8 4 66346 9.5 >7.8 and ≤8.4 5 61272 8.9 >8.4 and ≤9.1 6 68823 9.9 >9.1 and ≤10.4 7 68464 9.9 >10.4 and ≤14.9 8 66292 9.5 >14.9 9 75616 10.9
下載: 導出CSV
表 3 CNN?LSTM模型參數設置
Table 3. Parameter setting of the CNN?LSTM model
Layers Type Neurons Filters Kernel-size Strides Padding Pool-size 1 Conv1d (1,1,700) 8 3 1 0 — 2 BatchNorm1d (8,1, 698) — — — — — 3 ReLU (8,1, 698) — — — — — 4 MaxPool1d (8,1, 698) — — — 0 2 5 Conv1d (8,1, 348) 16 5 1 0 — 6 BatchNorm1d (16,1, 344) — — — — — 7 ReLU (16,1, 344) — — — — — 8 MaxPool1d (16,1, 344) — — — 0 2 9 Conv1d (16,1, 172) 32 8 1 0 — 10 BatchNorm1d (32,1, 165) — — — — — 11 ReLU (32,1, 165) — — — — — 12 MaxPool1d (32,1,165) — — — 0 2 13 Conv1d (32,1, 83) 128 2 1 0 ? 14 BatchNorm1d (128,1, 82) — — — — — 15 ReLU (128,1, 82) — — — — — 16 LSTM (128,1, 82) 128 — — — — 17 Fully-connected (1,128) — — — — — 18 Fully-connected (1, 64) — — — — — 19 Output 10 — — — — —
下載: 導出CSV
表 4 A1、A2、B1和B2分別進行個體建模性能評估
Table 4. Individual modeling performance evaluations for A1, A2, B1, and B2
Volunteer Precision Recall F1-score A1 0.79 0.79 0.79 A2 0.80 0.80 0.80 B1 0.81 0.79 0.79 B2 0.86 0.86 0.86
下載: 導出CSV
表 5 群體建模下的血糖監測混淆矩陣
Table 5. Confusion matrix for blood glucose prediction under group modeling
Labels Predict 0 Predict 1 Predict 2 Predict 3 Predict 4 Predict 5 Predict 6 Predict 7 Predict 8 Predict 9 0 1083 45 9 10 0 0 0 0 0 0 1 37 1021 35 29 0 0 0 0 0 0 2 9 35 1093 31 0 1 0 0 0 0 3 10 90 98 926 34 0 0 0 0 0 4 0 0 0 20 1119 47 12 1 1 2 5 0 0 1 0 65 945 77 2 8 40 6 0 0 0 0 3 4 1010 99 24 50 7 0 0 0 0 2 0 19 966 87 71 8 0 0 0 0 2 0 6 43 982 105 9 0 0 0 0 5 10 24 3 27 1051
下載: 導出CSV
表 6 血糖監測模型性能評估
Table 6. Performance evaluation of the proposed glucose prediction model
Labels Precision Recall F1-score 0 0.95 0.94 0.95 1 0.86 0.91 0.88 2 0.88 0.93 0.91 3 0.91 0.80 0.85 4 0.91 0.93 0.92 5 0.94 0.83 0.88 6 0.88 0.85 0.86 7 0.87 0.84 0.86 8 0.87 0.86 0.87 9 0.80 0.94 0.86
下載: 導出CSV
表 7 血糖監測模型對比
Table 7. Comparison of glucose prediction models
Related work Classification Using signals Modeling method Model Precision/% Literature[16] 6 ECG+PPG Individual modeling ELM 83.5 CNN 81.2 Fractional order system 77.3 This paper 10 ECG Individual modeling CNN?LSTM 81.5 Group modeling CNN?LSTM 88.4
下載: 導出CSV
www.77susu.com<span id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> <span id="fpn9h"><noframes id="fpn9h"> <th id="fpn9h"></th> <strike id="fpn9h"><noframes id="fpn9h"><strike id="fpn9h"></strike> <th id="fpn9h"><noframes id="fpn9h"> <span id="fpn9h"><video id="fpn9h"></video></span> <ruby id="fpn9h"></ruby> <strike id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> -
參考文獻
[1] Wan X S. Review of the risk factors and intervention of type 2 diabetes. Chin J Soc Med, 2006, 23(4): 251 doi: 10.3969/j.issn.1673-5625.2006.04.019萬曉珊. 2型糖尿病的危險因素及干預綜述. 中國社會醫學雜志, 2006, 23(4):251 doi: 10.3969/j.issn.1673-5625.2006.04.019 [2] Zheng C Z, Ding D. Guiding opinions of Chinese diabetes surgery experts (2010). Chin J Pract Surg, 2011, 31(1): 54鄭成竹, 丁丹. 中國糖尿病外科治療專家指導意見(2010). 中國實用外科雜志, 2011, 31(1):54 [3] Hou Q T, Li Y, Li S Y, et al. The global burden of diabetes mellitus. Chin J Diabetes, 2016, 24(1): 92 doi: 10.3969/j.issn.1006-6187.2016.01.023侯清濤, 李蕓, 李舍予, 等. 全球糖尿病疾病負擔現狀. 中國糖尿病雜志, 2016, 24(1):92 doi: 10.3969/j.issn.1006-6187.2016.01.023 [4] Zheng Z J. Application of Continuous Glucose Monitoring System in Critical Illness: A Preliminary Study [Dissertation]. Hangzhou: Zhejiang University, 2016鄭忠駿. 連續血糖監測系統在危重患者中應用的初步研究[學位論文]. 杭州: 浙江大學, 2016. [5] Wei Z, Zhang B X, Shi H B, et al. The development of noninvasive detection technique of blood glucose. China Med Equip, 2020, 17(12): 196 doi: 10.3969/J.ISSN.1672-8270.2020.12.046韋哲, 張秉璽, 石恒兵, 等. 無創血糖檢測技術的發展. 中國醫學裝備, 2020, 17(12):196 doi: 10.3969/J.ISSN.1672-8270.2020.12.046 [6] Wang L P. Study on Approach of ECG Classification with Domain Knowledge [Dissertation]. Shanghai: East China Normal University, 2013王麗蘋. 融合領域知識的心電圖分類方法研究[學位論文]. 上海: 華東師范大學, 2013. [7] Liu C Y, Yang M C, Di J N, et al. Wearable ECG: History, key technologies and future challenges. Chin J Biomed Eng, 2019, 38(6): 641 doi: 10.3969/j.issn.0258-8021.2019.06.001劉澄玉, 楊美程, 邸佳楠, 等. 穿戴式心電: 發展歷程、核心技術與未來挑戰. 中國生物醫學工程學報, 2019, 38(6):641 doi: 10.3969/j.issn.0258-8021.2019.06.001 [8] Wang J H. What are the hazards of hyperglycemia and hypoglycemia. Prevent treat cardiovasc, 2016(4): 19王建華. 高血糖、低血糖各有哪些危害. 心血管病防治知識, 2016(4):19 [9] Acharya U R, Fujita H, Lih O S, et al. Automated detection of coronary artery disease using different durations of ECG segments with convolutional neural network. Knowl Based Syst, 2017, 132: 62 doi: 10.1016/j.knosys.2017.06.003 [10] Pfeifer M A, Cook D, Brodsky J, et al. Quantitative evaluation of cardiac parasympathetic activity in normal and diabetic man. Diabetes, 1982, 31(4pt1): 339 [11] Porumb M, Stranges S, Pescapè A, et al. Precision medicine and artificial intelligence: A pilot study on deep learning for hypoglycemic events detection based on ECG. Sci Rep, 2020, 10: 170 doi: 10.1038/s41598-019-56927-5 [12] Jin X F, Liu C H, Xu T L, et al. Artificial intelligence biosensors: Challenges and prospects. Biosens Bioelectron, 2020, 165: 112412 doi: 10.1016/j.bios.2020.112412 [13] Tobore I, Kandwal A, Li J Z, et al. Towards adequate prediction of prediabetes using spatiotemporal ECG and EEG feature analysis and weight-based multi-model approach. Knowl Based Syst, 2020, 209: 106464 doi: 10.1016/j.knosys.2020.106464 [14] Kandhasamy J P, Balamurali S. Performance analysis of classifier models to predict diabetes mellitus. Procedia Comput Sci, 2015, 47: 45 doi: 10.1016/j.procs.2015.03.182 [15] Tafa Z, Pervetica N, Karahoda B. An intelligent system for diabetes prediction//2015 4th Mediterranean Conference on Embedded Computing (MECO). Budva, 2015: 378 [16] Liu Y W. Research on Non-Invasive Blood Glucose Detection Based on PPG and ECG Fusion Signal [Dissertation]. Guangzhou: Guangdong University of Technology, 2018劉宇巍. 基于PPG和ECG信號融合的無創血糖檢測方法研究[學位論文]. 廣州: 廣東工業大學, 2018. [17] Feng P H. Research on Blood Glucose Estimation Based on Signal Processing Technology [Dissertation]. Guangzhou: Guangdong University of Technology, 2019馮培華. 基于信號處理技術的無創血糖估計研究[學位論文]. 廣州: 廣東工業大學, 2019. [18] Tobore I, Li J, Yuhang L, et al. Deep learning intervention for health care challenges: Some biomedical domain considerations. JMIR Mhealth Uhealth, 2019, 7(8): e11966 doi: 10.2196/11966 [19] Tobore I, Li J, Kandwal A, et al. Statistical and spectral analysis of ECG signal towards achieving non-invasive blood glucose monitoring. BMC Med Inform Decis Mak, 2019, 19(Suppl6): 266 [20] Chen W, Xie X S, Wang J L, et al. A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility. CATENA, 2017, 151: 147 doi: 10.1016/j.catena.2016.11.032 [21] Yang H B, Wu X W, Yuan R, et al. Establishment and verification of the prediction model and nomogram for type 2 diabetes blood glucose control [J/OL]. J Chengdu Med College (2020-7-22) [2020-12-30]. http://kns.cnki.net/kcms/detail/51.1705.R.20200722.1640.004.html.楊恒博, 吳行偉, 袁蓉, 等. 2型糖尿病血糖控制預測模型及列線圖的建立與驗證[J/OL]. 成都醫學院學報(2020-7-22) [2020-12-30]. http://kns.cnki.net/kcms/detail/51.1705.R.20200722.1640.004.html. [22] Teng J L, Rong Z J, Xu Y, et al. Blood glucose prediction method based on gru. Comput Appl Softw, 2020, 37(10): 107 doi: 10.3969/j.issn.1000-386x.2020.10.017滕建麗, 容芷君, 許瑩, 等. 基于GRU網絡的血糖預測方法研究. 計算機應用與軟件, 2020, 37(10):107 doi: 10.3969/j.issn.1000-386x.2020.10.017 [23] Seyd A P T, Joseph P K, Jacob J. Automated diagnosis of diabetes using heart rate variability signals. J Med Syst, 2012, 36(3): 1935 doi: 10.1007/s10916-011-9653-x [24] Acharya U R, Fujita H, Oh S L, et al. Application of deep convolutional neural network for automated detection of myocardial infarction using ECG signals. Inf Sci, 2017, 415-416: 190 doi: 10.1016/j.ins.2017.06.027 [25] Ashiquzzaman A, Tushar A K, Islam M R, et al. Reduction of overfitting in diabetes prediction using deep learning neural network//IT Convergence and Security 2017. Singapore, 2017: 35 [26] Swapna G, Acharya R U, Vinithasree S, et al. Automated detection of diabetes using higher order spectral features extracted from heart rate signals. Intell Data Anal, 2013, 17(2): 309 doi: 10.3233/IDA-130580 [27] Porumb M, Griffen C, Hattersley J, et al. Nocturnal low glucose detection in healthy elderly from one-lead ECG using convolutional denoising autoencoders. Biomed Signal Process Control, 2020, 62: 102054 doi: 10.1016/j.bspc.2020.102054 [28] Yannakoulia M, Lykou A, Kastorini C M, et al. Socio-economic and lifestyle parameters associated with diet quality of children and adolescents using classification and regression tree analysis: The DIATROFI study. Public Health Nutr, 2016, 19(2): 339 doi: 10.1017/S136898001500110X [29] Yuan C X, Jia D N, Zhou S H. Research and application of convolutional neural network in mining area prediction. Chin J Eng, 2020, 42(12): 1597袁傳新, 賈東寧, 周生輝. 卷積神經網絡在礦區預測中的研究與應用. 工程科學學報, 2020, 42(12):1597 -
點擊查看大圖
圖(4) / 表(7)
計量
- 文章訪問數: 1440
- HTML全文瀏覽量: 757
- PDF下載量: 153
- 被引次數: 0
