作者: S. Tsantis , N. Dimitropoulos , D. Cavouras , G. Nikiforidis
DOI: 10.1016/J.CMPB.2006.09.006
关键词: Boundary detection 、 Wavelet 、 Hough transform 、 Artificial intelligence 、 Computer science 、 Edge detection 、 Segmentation 、 Maxima and minima 、 Wavelet transform 、 Scale model 、 Computer vision 、 Region of interest
摘要: A hybrid model for thyroid nodule boundary detection on ultrasound images is introduced. The segmentation combines the advantages of ''a trous'' wavelet transform to detect sharp gray-level variations and efficiency Hough discriminate region interest within an environment with excessive structural noise. proposed method comprise three major steps: a edge procedure speckle reduction map estimation, based local maxima representation. Subsequently, multiscale structure utilised in order acquire contour representation by means chaining similar attributes form significant structures. Finally, employed 'a priori' knowledge related nodule's shape distinguish from adjacent comparative study between our automatic manual delineations demonstrated that boundaries extracted are closely correlated physicians. can be value nodules' shape-based classification as educational tool inexperienced radiologists.