On the truncation of time frequency distributions to improve the computational performance in the estimation of fundamental parameters of a Doppler ultrasound blood flow signal

Abstract

Accurate estimation of mean velocity and bandwidth of a Doppler Ultrasound blood flow signal, is essential to evaluate the success of a coronary artery graft surgery. The estimation accuracy and efficiency obtained depends on the spectral method used and some fundamental parameters such as frequency resolution and window length. However, some suitable methods are very complex and as a consequence computationally very intensive which could not be appropriate for real-time applications. This paper presents a series of studies using time frequency distributions to estimate mean frequency and RMS bandwidth, introducing a truncation index in the context of measuring blood flow in arteries. Results show that truncating a number of terms in the generalized auto-correlation function would introduce errors that, if carefully chosen, will not be significant. Furthermore, besides the reduction on calculating an important number of terms, the method offers the alternative of increasing the spectral resolution at no extra computational cost, both important elements to be considered in real-time applications.