Non-invasion Blood Pressure Measurement Based On The Fusion Of ECG And PPG Information Based On Deep Learning

The Blood pressure is the pressure applied to a unit area of the blood vessel wall by pumping blood into the blood vessel when the heart contracts.It is measured in millimeters of mercury(mm Hg)and is one of the important vital signs.Blood pressure can reflect cardiovascular health,physical function and other conditions,and plays a guiding role when we pay attention to our own physical health.With the acceleration of China's aging population process,health problems related to hypertension have become increasingly prominent.The current non-invasive blood pressure detection methods are limited by the application scenarios of blood pressure equipment and traditional embedded devices derived from pulse wave velocity.They have the disadvantages of low accuracy and the need of repeated calibration,which makes it difficult to meet the reality of normalized blood pressure measurement,far from the demand of portability and accuracy of the high blood pressure and potential patients with high blood pressure.In this paper,we will use deep learning-based algorithms to model multiple feature values extracted from Photoplethysmography(PPG)and electrocardiogram(ECG)signals,and design a model that is both long-term and highly accuracy non-invasive blood pressure measurement system.The hardware verification platform of the system is implemented based on the STM32 platform,which can simultaneously collect PPG and ECG signals,and upload the data to the computer for processing.The training data of the network model comes from the Medical Information Mart for Intensive Care(MIMIC).In this paper,a number of different algorithms are used to model the prediction of blood pressure respectively,and their advantages and disadvantages are compared.These blood pressure prediction models are based on the relationship between pulse wave propagation time and blood pressure,and the multi-feature fusion of PPG and ECG signals applied to machine learning and deep learning algorithms.The final comparative analysis concluded that the convolutional neural network(CNN)combined with sequence-to-sequence network(seq2seq)model has the best effect on continuous blood pressure prediction.The research in this paper combines knowledge in the fields of physiological signal detection,embedded software and hardware,and machine learning and deep learning.The PPG and ECG signals based on deep learning have mainly completed the following tasks in the continuous blood pressure detection system:1.Made a systematic overview of the research background of blood pressure measurement methods and the current development status at home and abroad.Analyzed the requirements of non-invasive blood pressure testing equipment,and gave specific design plans,researched and realized the traditional pulse wave transit time(PWTT)non-invasive blood pressure measurement method,enhanced learning,random forest and other machine learning algorithms,CNN and CNN +seq2seq deep learning algorithm for continuous blood pressure prediction.2.The hardware circuit platform for pulse wave measurement was realized,including PPG and ECG signal synchronous acquisition circuit,charging and voltage stabilization circuit,and the design ideas and circuit principles are introduced.3.The PWTT feature extraction algorithm was introduced in detail,focusing on extracting the ratio of pulse wave frequency to blood viscosity,photoplethysmography intensity ratio(PIR),and pulse arrival time(PAT)from PPG and ECG signals.We introduced the use of machine learning algorithms,network frameworks used in deep learning,and design reasons.4.Realized the modeling process of the PWTT algorithm and its correlation analysis,as well as the prediction of blood pressure by machine learning and deep learning models.The analysis of model prediction results proves that multi-feature fusion machine learning and deep learning algorithms are superior than traditional methods of calculating blood pressure based on pulse wave propagation time.It can use wireless networks to provide services for specific environments such as intensive care units.The precise non-invasive blood pressure equipment provides a powerful algorithm reserve.