A study of coronary intravascular ultrasound RF signals using wavelets

Attempts to automate analysis of Intravascular Ultrasound (IVUS) images of coronary arteries for tissue boundary and plaque identification have not proven clinically successful. Utilising the raw backscattered Radio Frequency (RF) IVUS signal which contains significantly more information than the processed image, we have compared a number of techniques for determining the position of the edges of vessel walls and plaques and have developed a wavelet-based technique that overcomes inherent problems with image analysis. Ten 40MHz RF signals were quadrature demodulated to preserve the phase information in the signal and the resulting real and imaginary components squared and added to produce a measure of instantaneous RF signal energy. Wavelet decompositions of up to five levels were then carried out using a number of different wavelets on each RF signal. The IVUS images were displayed as gray-scale images and the separate decompositions were then scaled and mapped as a colour intensity strip onto their corresponding brightness/intensity level to replicate image scan lines. The results showed that the energy to image intensity transformations at the second and third approximations and at the third and fourth details could adequately identify the blood-vessel boundaries. This was verified by superimposing the wavelet decomposition levels on the original IVUS images in conjunction with expert manual identification of the bloodtissue interface of the vessel. Good agreement was demonstrated between automated and manual results.