Research On Key Technologies In Cardiovascular Function Evaluation Based On Wearable Technology

Cardiovascular diseases(CVDs)have become a major problem that seriously endangers global public health,surpassing cancer as the leading cause of death.The situation of CVDs prevention and control in China is more serious,"China Cardiovascular Diseases Report 2018 Summary" pointed out that the number of cardiovascular patients in China has reached 290 million,the mortality rate of CVDs accounted for more than 40% of the total mortality rate.CVD is a typical chronic disease because of its many complications,long disease period and huge consumption of medical and health resources.With the release of "Healthy China Action",China is gradually changing from a "disease-centered" medical and health system to a "health-centered" medical and health system.Early detection and treatment of diseases are an effective means to reduce the probability of people's suffering from critical diseases.At present,the diagnosis and treatment technology of CVDs in hospital has become more and more mature,various types of CVDs have effective treatment methods.However,facing the cardiovascular health care in daily life,the traditional hospital treatment methods are insufficient.Therefore,for the early patients of CVDs and more subhealthy people,it is of great significance to explore the theory and practice of wearable cardiovascular health measurement and evaluation to realize the early screening and disease assessment of cardiovascular disease.Based on electrocardiogram(ECG)and photoplethysmography(PPG),this paper conducts a series of researches,such as assisted diagnosis of hypertension disease,new cuffless beat-to-beat blood pressure detection method and automatic identification classification of heart diseases.First of all,this study explores the hypertension auxiliary diagnosis method based on wearable PPG signal for the first time,this method only needs PPG signal which is low cost and easy access and provides a new solution to achieve the automatic identification of hypertension for wearable health devices.It will play an important role in the prevention and control of cardiovascular diseases;Secondly,this study develops and perfects the theory of arterial wave propagation and pulse wave morphology.The method based on ECG and PPG signal realizes cuff-less blood pressure measurement and meets the AAMI and BHS evaluation standards,which provides important theoretical support for novel wearable blood pressure detection.Finally,a CNN-Bi LSTM deep learning model is constructed based on 12-lead ECG signals which realizes the automatic identification and diagnosis of various heart diseases.Further,a comparison study with evolutionary-neural system approach is conducted based on single-lead ECG signal which shows the feasibility applied in wearable ECG device This study provides the theoretical and experimental support for wearable cardiovascular health devices and provides new solutions for early detection of cardiovascular disease.Main research works and results are included as below:(1)Clinical data collection experiment and signal processing research.Medical clinical data collection experiment is a standardized and demanding scientific research activities,subjects' privacy,safety risks and informed consent need to be implemented canonically,otherwise there will be many medical ethics issues.The study organized and obtained a BPPPG dataset containing 219 subjects by the medical clinical data collection programs.Based on this dataset,signal quality evaluation of PPG signal and the optimal filter are studied,and an effective method for denoising PPG signal is obtained.This study finds that the 4th-order II Chebyshev bandpass filter can significantly improve the signal quality of PPG.(2)A study of the auxiliary diagnosis of hypertension based on PPG signal.The traditional diagnosis of hypertension is determined by blood pressure,which is limited by the restriction of cuff blood pressure.With the development of wearable technology,it is necessary to be more convenient for the diagnosis and identification of hypertension in wearable devices.Based on the BP-PPG dataset and MIMIC dataset which is widely used,this study carried out the hypertension assisted diagnosis models with three different theoretical.This study finds that the method adopting to deep learning and PPG technology which achieves the 92.55% F1 score can replace the method using ECG and PPG to classify hypertension and normotension.It validates the feasibility of hypertension diagnosis using single PPG signal and provides a new solution for wearable cardiovascular health devices.(3)A study of new cuff-less beat-by-beat blood pressure detection method based on ECG and PPG.New cuff-less blood pressure detection technology has become an urgent need in current society.Based on arterial wave propagation theory and PPG morphology theory,this study explores the new cuff-less beat-by-beat blood pressure detection method.This study finds that the method based on ECG and PPG can meet AAMI and Grade A BHS blood pressure meter evaluation standard.It provides important theoretical support for the new wearable continuous blood pressure detection.(4)A study of heart diseases auxiliary diagnostic methods based on ECG signal.CVDs have become the world's leading cause of death,and heart disease often causes disabilities or death due to its suddenness and short time for treatment.This study constructed a convolution and Bidirectional long short-term memory network(CNN-Bi LSTM)and realized the automatic identification and classification of 9 types heartbeat events based on 12 leads ECG signals.In this study,the F1 score of RBBB reached 0.943,and overall F1 score reached 0.800.Further,the model was verified and compared with neural evolution system in the single lead ECG signal.It is feasible to be applied in wearable heart diseases assessment.