Dynamic Programming and Fuzzy Classification for the Automatic Segmentation of the Carotid in Ultrasound Images

A new approach is proposed for the automatic detection of the near-end and far-end intima and adventitia inner boundaries in ultrasound images of the common carotid artery. This method uses the instantaneous coefficient of variation edge detector, fuzzy classification of edges, several discriminating features of the carotid wall boundaries and dynamic programming. The carotid wall boundaries are detected both in healthy and in atherosclerotic arteries, with a wide range of plaque types and sizes. Manual and automatic results are significantly better for the far-end wall, where the automatic detection shows an accuracy similar to manual detections. The application of this approach in clinical practice is encouraged by the results for the far-end wall and the short computation time.