Abstract:Background: Cardiovascular disease is one of the leading causes of death worldwide. However, according to studies, 90% of cardiovascular diseases can be prevented. Cardiovascular function parameters are an important basis for the diagnosis of cardiovascular diseases. The pulse wave also contains a wealth of physiological and pathological information, which can reflect the trend of cardiac function parameters at an early stage, so the measurement and analysis of the pulse wave signal becomes more and more important. The wearable pulse signal acquisition device has gradually become a new trend. In the mobile health scenario, convenient use is the prerequisite for long-term and rapid health monitoring. The data containing diverse pulse wave signals is the basis for obtaining more comprehensive and accurate human physiopathological information. Accurate data analysis and processing is the key to realizing the important goal of cardiovascular health monitoring. Objective: Based on the concept of mobile health care, wearable devices are developed to obtain physiological signals. The zero-dimensional model and the optimization algorithm are combined to complete the uncertainty quantification of the microcirculation parameters. Then, a feature set containing the cardiovasvular parameters can be constructed. The machine learning algorithm can be used to build a model that can accurately realize cardiovascular disease identification. Methods: This paper adopts laboratory-developed equipment to acquire the wrist pulse wave and fingertip volume pulse wave. A total of 323 samples were collected from healthy populations, hypertensive patients and patients with coronary heart disease (CHD). The pulse blood flow model in fingertip microcirculation is established, and the uncertainty quantification of model parameters is completed based on slime mold algorithm (SMA). After comparing and analyzing the performance of four algorithms on pulse wave classification, the identification model of cardiovascular diseases is established based on the microcirculatory characteristic parameter set and random forest algorithm (RF). Results: RF showed good classification performance among the four classification algorithms. The identification accuracy of the model built on the microcirculatory characteristic parameter set and RF algorithm all reached more than 88%. The highest recognition accuracy was 95.51% for coronary heart disease samples, 92.11% for healthy samples, and 88.55% for hypertensive samples. It can be seen that the model based on RF algorithm has a good ability to distinguish the characteristic parameters in different cardiovascular health states. Conclusions: The wearable device designed in this paper can facilitate the daily health monitoring of cardiovascular disease. By using the combination of the physical model and machine learning model, the uncertainty quantification of microcirculation parameters and the identification of cardiovascular disease was finally completed. The recognition model based on machine learning provides a new idea and method for the research of cardiovascular health monitoring through pulse waves.

A Novel Wearable Multi-Channel Pulse Monitor Combined with Machine Learning for Screening Hypertension Population

Longitudinal Wrist PPG Analysis for Reliable Hypertension Risk Screening Using Deep Learning

A new method of continuous blood pressure monitoring using multichannel sensing signals on the wrist

Uncertainty quantification of microcirculatory characteristic parameters for recognition of cardiovascular diseases

Multi-Dimensional Feature Combination Method for Continuous Blood Pressure Measurement Based on Wrist PPG Sensor.

Wearable multichannel-active pressurized pulse sensing platform

A Novel Single-Arm-worn 24 H Heart Disease Monitor Empowered by Machine Intelligence

A Wearable ECG Monitor for Deep Learning Based Real-Time Cardiovascular Disease Detection

Prediction of hypertension risk based on multiple feature fusion

Synchronized wearables for the detection of haemodynamic states via electrocardiography and multispectral photoplethysmography

Cuffless And Continuous Blood Pressure Estimation From The Heart Sound Signals

Monitoring blood pressure and cardiac function without positioning via a deep learning–assisted strain sensor array

Recognition of Pulse Wave Feature Points and Non-invasive Blood Pressure Measurement

Wearable cuffless blood pressure monitoring: from flexible electronics to machine learning

Photoplethysmography and Deep Learning: Enhancing Hypertension Risk Stratification

[A non-contact continuous blood pressure measurement method based on video stream]

Wearable multichannel pulse condition monitoring system based on flexible pressure sensor arrays

Ambulatory Cardiovascular Monitoring Via a Machine‐Learning‐Assisted Textile Triboelectric Sensor

Multi-kernel Convolutional Neural Network for Wrist Pulse Signal Classification

Blood pressure monitoring via double sandwich-structured triboelectric sensors and deep learning models

A survey: From shallow to deep machine learning approaches for blood pressure estimation using biosensors