Blood pressure prediction method based on feature fusion

Abstract

The invention discloses a blood pressure prediction method based on feature fusion. The method comprises the steps that original pulse wave signals are collected; the original pulse wave signals are sequentially subjected to preprocessing such as filtering, baseline drift removal and single-cycle extraction, so that processed pulse wave signals are obtained; feature extraction is conducted on the pulse wave signals processed in the step 2 to obtain time domain features of the preprocessed pulse wave signals, amplitude change features of pressure pulse waves under different levels of pressure and volume pulse wave conduction speed, and screening and fusion of the features are completed based on an embedded feature selection method, so that a feature set used for blood pressure prediction is obtained; the feature set is trained based on a random forest regression algorithm to obtain SBP and DBP prediction models; and the blood pressure is predicted by using the SBP and DBP prediction models. According to the method, accurate and effective prediction of SBP and DBP is achieved, the AAMI use standard can be met, and the method has the advantage of being accurate in prediction.

Description

technical field [0001] The present invention relates to the technical field of blood pressure prediction methods, in particular to a blood pressure prediction method based on feature fusion. Background technique [0002] Arterial blood pressure is an important physiological indicator for judging the health status of the human cardiovascular system, and it also plays an important role in the analysis and identification of pulse signals. Recently, non-invasive blood pressure measurement technology has continuously improved with the improvement of scientific and technological level, and has made great progress in clinical application. Compared with the invasive type, the non-invasive type has higher safety and convenience, and is suitable for the application of wearable devices. [0003] Currently, there are two main types of non-invasive blood pressure prediction methods based on pulse waves[84]. (Diastolic Blood Pressure, DBP), the other is to extract the information carrie...

AI Technical Summary

Problems solved by technology

The first type of method can obtain more accurate blood pressure values, but usually requires continuous pressure on the upper limbs, the equipment is complex, time-consuming, and may cause injury if used for a long time; the second type of measurement method is more convenient and suitable for rapid blood pressure measurement Medium, but it is necessary to obtain features that are strongly correlated with blood pressure to improve the accuracy of blood pressure prediction

Images