Research On Noninvasive Blood Pressure Prediction Method Of Morphological Characteristics Of Pulse Wave Fusion Conduction Time

Blood pressure as an important physiological index of human body,its numerical changes can objectively reflect the cardiovascular health of human body,accurate access to real-time human blood pressure values,has an important guiding significance for clinical diagnosis of modern medicine.Blood pressure,as an important basis for disease diagnosis,can be monitored in real time and continuously,which is of great strategic significance for the prevention and control of cardiovascular diseases.With the rapid development of blood pressure measurement technology in recent years,many accurate measurement methods have emerged,but most of them do not have the ability of continuous real-time measurement.Arterial intubation method is known as the international gold standard of blood pressure measurement,which can achieve continuous and accurate measurement of blood pressure.However,the measurement process of arterial intubation method will cause certain injuries to human body,and the technical requirements are high.To sum up,modern blood pressure measurement technology is developing in three directions:continuous measurement,non-invasive measurement and accurate measurement.Pulse wave has the characteristics of convenient acquisition,which can well meet the two conditions of non-invasive and continuous.Moreover,with the further development of intelligent algorithm,it is expected that the measurement accuracy can be improved to some extent.Therefore,the method of blood pressure measurement based on pulse wave signal has gradually become a research hotspot in this field.The methods of blood pressure measurement based on pulse wave can be divided into two categories:conduction time-based and morphological characteristics-based.The conduction time has the characteristics of convenient extraction and good correlation,but the design of acquisition circuit is relatively complicated.Although the acquisition process of morphological features is simple,the signal morphology difference is great,the extraction is complex,and the correlation with blood pressure is low.Based on the open medical data set,this paper attempts to integrate the advantages of the two,modeling by parameter fusion,and evaluates the model performance.Specific work contents and achievements are as follows:The research work in this paper is based on MIMIC database.In order to build a high quality experimental data set,this paper designs an effective signal denoising and BP extraction scheme,and proposes a dynamic threshold feature point recognition algorithm.Based on the constructed data set,various correlations between conduction time and blood pressure were analyzed in a single sample.Based on the experimental results,an optimal conduction time selection method based on TOPSIS comprehensive evaluation method was further proposed.The experimental verification shows that compared with the traditional conduction time calculation method,the prediction MAE based on the optimal conduction time model is reduced by 0.21mmHg and the accuracy rate is increased by 2.75%,which can achieve better prediction effect.Based on the feature points identified by the above method,the morphological features were calculated,and the conduction time was fused with the morphological feature parameters to construct a linear regression model,a variety of nonlinear machine learning models and an ensemble learning model,respectively,which were evaluated on the test set.Through error analysis,compared with the model based on morphological characteristics,the model combining conduction time reduced the predicted MAE of systolic and diastolic blood pressure by 25.5%and 23.5%,respectively,and the performance was better improved.With the introduction of the integrated learning model,MAE prediction was further reduced by 0.18 mmHg,and the accuracy was improved by 2.8%,thus realizing more accurate blood pressure prediction.