Processing of CW Doppler images to extract velocity profile

The present work aims to find a good way to automatically trace the velocity profile in vessels shown in continuous wave Doppler spectrograms, replacing the more traditional manual tracing, easy to perform but rich in drawbacks. Different methods of pre-processing this kind of images in order to prepare the edge detection step are presented. Various techniques, taken from the literature or newly created, are tested on a set of 18 CW Doppler spectrograms. The main purpose is to understand which is the best strategy to put into practice, remembering that the goal is to get the maximum velocity envelope, or the mean velocity in the vessel. Two main strategies are tested: a mild pre-processing before edge detection, followed by an edge linking step to fill the gaps in the extracted contour, or a stronger pre-processing that guarantees a continuous contour without further operations. A comparison is made on the obtained results and it is shown that the two approaches are somehow complementary in their pros and cons. In further work, the strengths of the two strategies should be combined in a hybrid method that would guarantee a good compromise between continuity of edges and mathematical-based detection of boundaries