An Interleaved Data Acquisition to Reduce Common Noise in Coronary Doppler Vibrometry

Abstract

The coronary Doppler vibrometry (CDV) is a non-invasive ultrasonic diagnosis method to detect vibrations on a heart wall caused by stenosis in the coronary artery. However, in previous studies, CDV has a critical problem that a part of heart wall adjacent to normal blood vessel can also show high-frequency motion caused by common noise. To remove this noise, we obtain the ultrasound data at two heart wall segment locations at the same time. We transmit pulse to two different positions in an interleaved manner and obtain differential information of two reflected signals in order to eliminate common noise thereby enhancing the quality of stenosis-induced vibration. We confirmed the feasibility of the proposed approach through a clinical study. We beamformed acquired data from subjects and calculated wall velocity. To characterize the vibration, we generated the velocity spectrogram of heart wall by using short-time Fourier Transform. Then, to remove common noise, we extracted common signal from two interleaved data and subtracted common signal from original one. After that, we estimated the vibration power by integrating the spectrogram intensity between 300 and 600 Hz in the early diastole period. We judged that patients have higher differential vibration power in two interleaved data than normal subjects. We obtained a sensitivity of 83% and a specificity of 80% in the right coronary artery (RCA) cases and obtained a sensitivity of 100% and a specificity of 83% in the left anterior descending (LAD) cases. Our experiments show the proposed interleaved data acquisition gives a potential of differentiating between normal and abnormal subjects. We also need to mention that there are still unresolved issues for further improvements such as low SNR from remaining noise, diagnostic difficulty in setting range gate, etc.