Accurate detection of coronary arteries with the continuous wavelet transform

The authors propose a multi-scale detection scheme of the vessels borders by means of the continuous wavelet transform. The Canny's (1986) performance parameters are derived for Mallat's (1992) wavelet based edge detection operator. The authors take the scale dependent behaviour of these parameters into account to design an algorithm that optimally combines the multiscale information. Evaluation of the algorithm accuracy is obtained on 4 vessel phantoms of different diameters. The authors compare the computer determined diameters to the actual diameters, the correlation coefficient is in the range 0.93-0.99 for different types of wavelets, which proves the ability of the method to accurately detect edges in the presence of noise.