Fuzzy C-Means Clustering for Myocardial Ischemia Estimation With Pulse Waveform Analysis

The purpose of this study is to build an automatic disease classification algorithm by pulse waveform analysis, based on a Fuzzy C-means clustering algorithm. A self designed three-axis mechanism was used to detect the optimal position to accurately measure the pressure pulse waveform (PPW). Considering the artery as a cylinder, the sensor should detect the PPW with the lowest possible distortion, and hence an analysis of the vascular geometry and an arterial model were used to design a standard positioning procedure based on the arterial diameter changed waveform for the X-axes (perpendicular to the forearm) and Z-axes (perpendicular to the radial artery). A fuzzy C-means algorithm was used to estimate the myocardial ischemia symptoms in 35 elderly subjects with the PPW of the radial artery. Two type parameters were used to make the features, one was a harmonic value of Fourier transfer, and the other was a form factor value. A receiver operating characteristics curve was used to determine the optimal decision function. The harmonic feature vector contain second, third and fourth harmonics ( H 2 , H 3 , H 4 ) performed at the level of 69% for sensitivity and 100% for specificity while the form factor feature vector derived from left hand (LFF) and right hand (RFF) performed at the level of 100% for sensitivity and 53% for specificity. The FCM- and ROC-based clustering approach may become an efficient alternative for distinguishing patients in the risk of myocardial ischemia, besides the traditional exercise ECG examination.