北京生物醫(yī)學(xué)工程

基于梯度提升樹的ECG-SAS自動識別方法

Automatic identification of ECG-SAS based on gradient boosting decision treealgorithm

作者：王偉梁杰牛洋洋劉洪濤

單位：深圳和而泰家居在線網(wǎng)絡(luò)科技有限公司(廣東深圳 ;518063) ; 深圳和而泰智能控制股份有限公司(廣東深圳 ;518057)

關(guān)鍵詞：睡眠呼吸暫停綜合征；梯度提升樹；心電圖；心率變異性；呼吸暫停低通氣指數(shù)

分類號：R318.04

出版年·卷·期（頁碼）：2019·38·6（617-622）

摘要：

目的睡眠呼吸暫停綜合征(sleep apnea syndrome, SAS)是威脅生命健康的多發(fā)病之一，目前判斷SAS的方法大多采用多導(dǎo)睡眠圖(polysomnography, PSG)，但其操作難度大、專業(yè)性高，不能有效推廣，因此，設(shè)計一種自動檢測SAS的方法顯得尤為迫切和重要。方法本文設(shè)計了一種基于梯度提升樹(gradient boosting decision tree, GBDT)的算法方案，首先通過信號處理方法提取心電圖（electrocardiogram，ECG）數(shù)據(jù)的心率變異性(heart rate variability, HRV)特征，然后結(jié)合上下文相關(guān)性策略處理HRV數(shù)據(jù)訓(xùn)練模型。在得到模型后，采用動態(tài)閾值策略微調(diào)預(yù)測結(jié)果。最后統(tǒng)計每小時內(nèi)的SAS發(fā)生次數(shù)，得到呼吸暫停低通氣指數(shù)(apnea–hypopnea index, AHI)，完成SAS病情預(yù)測。結(jié)果本文使用Apnea-ECG數(shù)據(jù)庫的ECG數(shù)據(jù)驗證該算法效果。結(jié)果顯示，采用本文方案，35個測試樣本的SAS單分鐘識別率為88.56%，按照AHI指標(biāo)，將樣本分為健康、輕度、中度、重度4類，其準(zhǔn)確率為91.43%。結(jié)論本文所述基于GBDT的SAS-ECG識別方案，可以有效檢測SAS事件，評估個體的SAS病情。

Objective Sleep apnea syndrome (SAS) is one of the most common life threatening diseases. At present, the method of judging SAS is polysomnography (PSG), with difficult and professional operation. This make it cannot be effectively popularized. Therefore, it is urgent and important to design an automatic SAS detection method. Methods In this paper, we designed a kind of ascension based on gradient boosting decision tree (GBDT) algorithm. In the scheme design, heart rate variability characteristics of electrocardiogram data were firstly extracted by signal processing method. Then we trained the model by using the HRV data which was processed according to the context correlation strategy, and used dynamic threshold strategy to fine-tune the prediction results. Finally, we counted the number of SAS case per hour, obtained the apnea-hypopnea index, and completed the SAS prediction. Results This paper used ECG data from Apnea-ECG database to verify the effectiveness of the algorithm. The SAS single-minute recognition rate of the 35 test samples was 88.56%, and the accuracy was 91.43% when samples were divided into four categories: healthy, mild, moderate and severe. Conclusions The ECG-SAS recognition scheme based on GBDT described in this paper can effectively detect SAS events and evaluate individual SAS conditions.

參考文獻(xiàn)：

[1] George CF. Sleep ? 5: Driving and automobile crashes in patients with obstructive sleep apnoea/hypopnoea syndrome[J]. Thorax, 2004, 59(9):804-807.

[2] Woods CE, Usher K, Maguire GP. Obstructive sleep apnoea in adult indigenous populations in high-income countries: an integrative review[J]. Sleep & Breathing, 2015, 19(1): 45-53.

[3] Redline S, Yenokyan G, Gottlieb DJ, et al. Obstructive sleep apnea-hypopnea and incident stroke: the sleep heart health study[J]. American Journal of Respiratory and Critical Care Medicine, 2010, 182(2): 269-277.

[4] Acharya UR, Chua EC, Faust O, et al. Automated detection of sleep apnea from electrocardiogram signals using nonlinear parameters[J]. Physiological Measurement, 2011, 32(3):287-303.

[5] Lweesy K, Fraiwan L, Khasawneh N, et al. New automated detection method of OSA based on artificial neural networks using P-wave shape and time changes[J]. Journal of Medical Systems, 2011, 35(4):723-734.

[6] Kesper K, Canisius S, Penzel T, et al. ECG signal analysis for the assessment of sleep-disordered breathing and sleep pattern[J]. Medical & Biological Engineering & Computing, 2012, 50(2):135-144.

[7] Guo D, Peng CK, Wu HL, et al. ECG-derived cardiopulmonary analysis of pediatric sleep-disordered breathing[J]. Sleep Medicine, 2011, 12(4): 384-389.

[8] Zarei A, Asl BM. Automatic detection of obstructive sleep apnea using wavelet transform and entropy based features from single-lead ECG signal[J]. IEEE Journal of Biomedical and Health Informatics, 2018, PP(99):1-1.

[9] Friedman JH. Greedy function approximation: a gradient boosting machine[J]. Annals of Statistics, 2001, 29 (5): 1189-1232.

[10] Chen T, Guestrin C. XGBoost: a scalable tree boosting system[C]// ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco, USA: ACM, 2016: 785-794.

[11] Penzel T, Moody GB, Mark RG, et al. The apnea-ECG database[C]// Computers in Cardiology. Cambridge MA USA, IEEE, 2000, 27: 255-258.

[12]Berntson GG, Bigger JTJr, Eckberg DL, et al. Heart rate variability: origins, methods, and interpretive caveats[J]. Psychophysiology, 2010, 34(6):623-648.

服務(wù)與反饋：

【文章下載】【加入收藏】

提示：您還未登錄，請登錄！點此登錄

51黑料吃瓜在线观看,51黑料官网|51黑料捷克街头搭讪_51黑料入口最新视频