心率變異預測心跳停止後甦醒之癲癇發作

Heart Rate Variability-based Seizure Prediction in Patients Survived from Cardiac Arrest

呂季桓(Ji-Huan Lyu)

指導教授 ： 趙福杉

共同指導教授 ： 謝建興(Jiann-Shing Shieh)

國立臺灣大學/工學院/醫學工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201901185

摘要

到院前心跳停止病人急救後復甦的預後非常差，原因是在心跳停止時長時間的缺氧，導致身體許多機能受損。其中腦細胞的受損將可能引發癲癇發作，及早並適當的對癲癇發作的病患做醫療處置，對於其預後將有很大的改善。腦電圖為診斷癲癇的最有效利器，然而在急診照護環境下長時間監測腦電圖十分不易，因此探索出一個相關的其他生理訊號來預測或早期偵測將具有高度的臨床及學術意義。

近年來許多文獻指出，心率變異度可作為早期癲癇發作偵測的指標，因此本研究於國立臺灣大學醫學院附設醫院接收15位到院前心跳停止後復甦的病人，各記錄將近72小時的心電圖及腦電圖，對於所記錄的訊號進行追溯研究，分析癲癇發作及未發作時的心率變異指標，利用支持向量機進行分類，建立模型，並利用交叉驗證法檢驗其準確度。

本研究共分析27筆樣本，以SDNN、pHF、LF/HF、sample entropy四樣心率變異參數所建立之模型表現最好，其靈敏度為66.7%，特異度為83.3%，預測正確率為77.8%。本研究亦建立一套完整的樣本蒐集與分析的流程，提供未來擴展樣本數，增進此預測模型的準確度。

關鍵字

到院前心跳停止 ； 癲癇 ； 心率變異分析 ； 支持向量機

並列摘要

The prognosis of patients who experience out-of-hospital cardiac arrest is poor because long-term hypoxia caused by cardiac arrest results in physical impairment. For instance, damage to brain cells may cause seizures. Early and appropriate medical treatment of patients with seizures improves their prognosis, and electroencephalography (EEG) is the most effective tool for diagnosing seizures. However, monitoring patients for a long time using EEG in an emergency care environment is challenging. Therefore, it is of considerable academic and clinical significance to identify another related physiological signal for early prediction or detection of seizures.

Many studies have reported that heart rate variability(HRV) can be used as an indicator for early seizure detection. Therefore, this study examined 15 patients who were admitted to National Taiwan University Hospital after out-of-hospital cardiac arrest, recorded nearly 72 hours of their electrocardiograms and electroencephalograms signals, and analyzed heart rate variability to identify a relationship between these samples and seizure occurrence. The results were classified using a support vector machine. A prediction model was established, and its accuracy was tested through cross-validation.

In this study, a total of 27 samples were analyzed. The model was established using four key HRV parameters: SDNN, pHF, LF/HF, and sample entropy. The sensitivity was 66.7%, specificity was 83.3%, and prediction accuracy was 77.8%. This study also established a complete process for sample collection and analysis, and if the number of samples can be expanded in the future, the accuracy of the prediction model can be improved.

並列關鍵字

OHCA ； seizure ； HRV ； support vector machine

參考文獻

[1] J. Berdowski, R. A. Berg, J. G. Tijssen, and R. W. Koster, "Global incidences of out-of-hospital cardiac arrest and survival rates: systematic review of 67 prospective studies," Resuscitation, vol. 81, no. 11, pp. 1479-1487, 2010.

[2] T. P. Aufderheide et al., "Resuscitation Outcomes Consortium (ROC) PRIMED cardiac arrest trial methods: Part 1: Rationale and methodology for the impedance threshold device (ITD) protocol," Resuscitation, vol. 78, no. 2, pp. 179-185, 2008.

[3] J. L. Bonnes et al., "Manual cardiopulmonary resuscitation versus CPR including a mechanical chest compression device in out-of-hospital cardiac arrest: a comprehensive meta-analysis from randomized and observational studies," Annals of emergency medicine, vol. 67, no. 3, pp. 349-360. e3, 2016.

[4] S. T. Youngquist, P. Ockerse, S. Hartsell, C. Stratford, and P. Taillac, "Mechanical chest compression devices are associated with poor neurological survival in a statewide registry: A propensity score analysis," Resuscitation, vol. 106, pp. 102-107, 2016.

[5] S. P. Keenan, P. Dodek, C. Martin, F. Priestap, M. Norena, and H. Wong, "Variation in length of intensive care unit stay after cardiac arrest: where you are is as important as who you are," Critical care medicine, vol. 35, no. 3, pp. 836-841, 2007.