A non-invasive continuous cuffless blood pressure estimation using dynamic Recurrent Neural Networks Umit Senturk a,⇑ , Kemal Polat b , Ibrahim Yucedag c a Department of Computer Engineering, Faculty of Technology, Duzce University, 81000 Duzce, Turkey b Department of Electrical and Electronics Engineering, Faculty of Engineering and Architecture, Bolu Abant Izzet Baysal University, 14280 Bolu, Turkey c Department of Computer Engineering, Faculty of Technology, Duzce University, 81000 Duzce, Turkey article info Article history: Received 26 January 2020 Received in revised form 21 May 2020 Accepted 8 July 2020 Keywords: NARX-NN RNN Cuffless blood pressure LSTM-NN Dynamic Neural Networks abstract Cardiovascular diseases (CVD) have become the most important health problem of our time. High blood pressure, which is cardiovascular disease, is a risk factor for death, stroke, and heart attack. Blood pres- sure measurement is commonly used to limit blood flow in the arm or wrist, with the cuff. Since blood pressure cannot be measured continuously in this method, the dynamics underlying blood pressure can- not be determined and are inefficient in capturing symptoms. This paper aims to perform blood pressure estimation using Photoplethysmography (PPG) and Electrocardiography (ECG) signals that do not obstruct the vascular access. These signals were filtered and segmented synchronously from the R inter- val of the ECG signal, and chaotic, time, and frequency domain features were subtracted, and estimation methods were applied. Different methods of machine learning in blood pressure estimation are com- pared. Dynamic learning methods such as Recurrent Neural Network (RNN), Nonlinear Autoregressive Network with Exogenous Inputs Neural Networks NARX-NN and Long-Short Term Memory Neural Network (LSTM-NN) used. Estimation results have been evaluated with performance criteria. Systolic Blood Pressure (SBP) error mean ± standard deviation = 0.0224 ± (2.211), Diastolic Blood Pressure (DBP) error mean ± standard deviation = 0.0417 ± (1.2193) values have been detected in NARX artificial neural network. The blood pressure estimation results are evaluated by the British Hypertension Society (BHS) and American National Standard for Medical Instrumentation ANSI/AAMI SP10: 2002. Finding the most accurate and easy method in blood pressure measurement will contribute to minimizing the errors. Ó 2020 Elsevier Ltd. All rights reserved. 1. Introduction Cardiovascular diseases (CVD) are among the leading diseases that cause human death today when the world population reaches 7.7 billion [1]. CVD’s primary risk factor is high blood pressure (BP), known as the silent killer. More than 10 million people are at high risk of BP [2–4]. This disease, which is mostly seen in the elderly population, has started to be seen in the lower age groups. Low physical activity [5], disorders in eating habits [6], increased con- sumption of animal fats [7–9] reduced the incidence of cardiovas- cular diseases to young and even children. BP, known as the pressure of blood on the vascular walls, can be neglected despite its easy measurement. Patients who are diagnosed with high BP try to keep BP within certain limits with the help of medication. If high BP diagnosis and treatment is not made in the early stages, it can have serious consequences such as heart attack, stroke, organ failure, and stroke [10–12]. Since BP is controlled by the sympa- thetic and parasympathetic nervous system, long-term blood pres- sure measurement is required in the diagnosis of the disease. BP devices used today are aneroid BP instruments that still use Korotr- off sounds. Appropriate device research is ongoing for long-term BP measurements. In the home and office environment, BP measurement vascular occlusion (VO) [13,14] and non-vascular occlusion (VNO) systems are used for BP measurement. In the VO blood pressure measure- ment systems, the sleeve that connects to the wrist or arm is filled with air and compresses the vessels and nerves. Therefore, they are not suitable for long-term blood pressure measurements. VNO measurement systems, on the other hand, are more suitable for long-term blood pressure measurements since they do not apply any pressure. Studies in the literature also tend to vascular access VNO blood pressure systems. Physiological signals such as Electro- cardiogram (ECG), Photoplethysmography (PPG) and Balistocardio- graphy (BCG) are used in systems that do not block the vascular https://doi.org/10.1016/j.apacoust.2020.107534 0003-682X/Ó 2020 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail addresses: usenturk98@gmail.com (U. Senturk), kpolat@ibu.edu.tr (K. Polat). Applied Acoustics 170 (2020) 107534 Contents lists available at ScienceDirect Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust