Research on central arterial pressure estimation algorithm based on clustering

Central arterial pressure (CAP) plays an important role in the detection and diagnosis of cardiovascular diseases. Although the traditional universal transfer function method has good accuracy in measuring central arterial pressure, it does not take into account the characteristic information contained in pulse wave shape, and the accuracy needs to be further improved.Therefore, a clustering based estimation algorithm for central arterial pressure is proposed. First of all, 5 layer Symlets wavelet and hard-soft compromised threshold was used to remove high-frequency noise of the data set. Secondly, in order to extract the characteristics of pulse waves themselves, k-means++ clustering was carried out for pulse waves in the training set. In order to reduce the over-fitting phenomenon, the initial number of cluster centers is selected as 1000. Finally, in each cluster, discrete Fourier transform is performed on pulse wave and central artery wave to obtain amplitude and phase data and train the transfer function. The test set was used for verification. Firstly, the corresponding clustering category of pulse wave was calculated, and then CAP was calculated by transfer function. The results showed that the absolute errors of systolic blood pressure (SBP), diastolic blood pressure (DBP) and pulse pressure (PP) were 2.50±2.22mmHg, 4.47±2.72mmHg and 4.60±3.73mmHg respectively. Compared with other algorithms, our method has better accuracy.