Artificial Neural Network Classification of Arterial Pulse Waveforms in Cardiovascular Diseases

Abstract

Diabetes mellitus and hypertension are common cardiovascular diseases nowadays. However, physical examinations for cardiovascular diseases are not available easily in clinical practice. The main purpose of this study is to extract and classify characteristics of arterial pulse waves in diabetics and hypertensive patients using artificial neural network. It is expected to be a useful way to identify unusual arterial pulse waveforms as a risk index for cardiovascular diseases. Continuous arterial pulse signal was acquired from 63 diabetics, 21 hypertensive patients and 25 healthy subjects during supine position. In this study, a self-organizing neural network system is established to classify the arterial pulse waveforms. The classifying system includes two stages. In the first stage, the arterial pulse signal was divided into segments for extracting characteristics to be the input pattern of neural network in the database. The second stage was to test unknown samples. Healthy arterial pulse templates were created based on these arterial pulse segments to show the unusual arterial pulse waveform tendency. Results revealed that there are 13 kinds of arterial pulse templates in healthy subjects’ database. Comparing with the arterial pulse waveforms templates in healthy subjects’ database, 73% diabetics and 90% hypertensive patients can be classified by the system. Moreover, 82% patients with severe diabetic autonomic neuropathy can be recognized. It can be speculated that cardiovascular diseases influence the function of autonomic nervous system to control blood pressure and changes the arterial pulse waveform. The results indicate unusual blood pressure waveforms could be distinguished based on healthy database via the neural network system. In conclusion, artificial neural network is an effective method to classify the arterial pulse waveform and it can be a fast way to detect the arterial pulse waveform out of ordinary.