Speaker: Preben Nas

Title: Fast multiscale edge-detection in medical ultrasound signals


The topic for the presentation is a fast space-scale filtering procedure and edge-detection algorithm for medical ultrasound signals. The edge-detector both detects and determines the precise location of major edges in such signals, without corrupting the representation with redundant information.

It is well-known that the continuous wavelet-transform and its maxima-lines effectively can be used to detect edges in noisy signals. Maxima-lines are obtained by computing the wavelet-transform at several close scales. However to reduce computational complexity and save time we suggest a time-scale filtering procedure which uses only a few distinct scales. The construction of the procedure is based on a study of maxima-lines corresponding to edges typical for medical ultrasound signals. This allows us to meet the demand of speed, but not on the expense of reliability.

The connection-procedure has been applied together with an edge-detector on several medical ultrasound signals. Our results show that we achieve good separation between noise and major edges in such signals.