The exact algorithm consists of the following steps:
1) Low pass filtering of the image (gaussian filter of the 50th order, standard deviation equal to 10) in order to increase signal-to-noise ratio.
2) Determination of the adventitial wall from the intensity profile of the image considered column wise. Specifically, for each column of the image, we identify the pixels possibly belonging to the adventitia as those local maxima whose intensity is in the 90th percentile of the intensity distribution. An example of a typical intensity profile derived along a column of the image is reported in Fig. 2.
3) Individuation of the carotid lumen as the minimum point on the intensity profile whose mean intensity and variance of the neighborhood belong to the lowest classes. The minimum is searched descending the intensity profile from the maxima found during step 2). When the minimum is reached, the corresponding point is taken as marker of the carotid lumen and the starting maximum as marker of the adventitia.
4) Iteration of steps 2) and 3) for all columns.