Seeking a Physiological Source of Noise Using Coinciding Patterns in a Multilead ECG Record

The noise is present in every real measurement data and specifically in recordings of the electrical activity of living tissue. In this case, due to a common physiological background it is particularly difficult to recognize the measured object and the unwanted activity of neighboring organs. This paper presents a method to discriminate the source of such physiological noise based on analysis of complex-valued time-frequency representation of a multilead record. The method uses a dual-tree pyramid decomposition and the polar transform to produce magnitude and phase values of high scale details. These details are inspected for coinciding patterns -the cross-correlation function separates the physiological source of noise with consistent phase shift and the circuitry noise with random phase shift. Based on magnitude ratio and phase delay, a triangular method indicates the most probable position of the noise source. The proposed method was implemented and tested with a simplified model of thorax and various real ECG records. Our results show that the position of noise source can be determined with an accuracy of order of 10% of the thorax diameter, what is sufficient to unambiguously indicate the active organ.