血壓監測是監測人們健康狀況的方法之一。異常血壓是各種心血管疾病的症狀。早期發現高血壓可以幫助患者早日治療並降低死亡率。因此，具有對患者的血壓變化進行實時監測的機制是非常有價值的。在本論文中，我們提出了一個基於心電圖（ECG）和脈搏傳導時間（PTT）的深度學習模型，用於連續估計收縮壓（SBP）和舒張壓（DBP）值。我們還進行了實驗來比較傳統機器學習方法和深度學習模型之間的準確性。在過去，許多研究人員已經嘗試在深度學習上進行殘餘連接以提高模型的性能。我們遵循這樣的做法並在每層LSTM中添加殘留物。對於實驗，我們使用Physionet的多參數智能監護在重症監護（MIMIC）II中的數據集作為心電圖（ECG）和光電容積描記圖（PPG）信號以及動脈血壓（ABP）信號的來源。我們使用了包括三個信號的12.000個主題。實驗結果表明，該方法可以將SBP的平均絕對誤差（MAE）值降至0.7357 mmHg，DBP降低0.5587 mmHg。我們還計算標準偏差（STD）值。 SBP的STD值為0.9579 mmHg，DBP的STD值為0.5088 mmHg。;Blood pressure monitoring is one of the ways to monitor people’s health conditions. The abnormal blood pressure is a symptom of various cardiovascular diseases. Early detection of hypertension can help patients get early treatment and reduce mortality. Therefore, it is very valuable to have a mechanism to perform real-time monitoring for the change of blood pressure in patients. In this thesis, we present a deep learning models based on electrocardiogram (ECG) and pulse transit time (PTT), for the continuous estimation of the systolic blood pressure (SBP) and diastolic blood pressure (DBP) values. We also performed experiments to compare the accuracy between the traditional machine learning method and deep learning model. In the past, many researchers already try residual connection on deep learning to increase the performance of the model. We followed such practice and added the residual inside each layer of LSTM. For the experiments, we used the dataset of Physionet’s multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II as the source of electrocardiogram (ECG) and photoplethysmogram (PPG) signals as well as the arterial blood pressure (ABP) signal. We used 12.000 subjects that include three signals. The experimental results show that the proposed method can reduce the value of mean absolute error (MAE) to 0.7357 mmHg for SBP and 0.5587 mmHg for DBP. We also calculate the standard deviation (STD) value. The STD values are 0.9579 mmHg for SBP and 0.5088 mmHg for DBP.
