Research On Extraction Of Medical Axes From Angiograms

Angiography is an effective technique for vessel visualization. Digital Subtraction Angiography (DSA for short) is a widely used technique developed from angiography. Axis extraction is the ground work for quantitative analysis and 3-D reconstruction of vessels. In this thesis, problems existing in angiographic vascular axis extraction were discussed, and an effective approach for axis extraction based on contextual enhancement was developed.We first employed the unsharp masking to adjust the contrast of angiograms, and then removed the consistent component of low frequency in frequency domain by high-pass filtering. An enhancement algorithm combining contextual information, which refers to the characteristic of biological visual mechanism and property of vascular structure, was brought forward in this thesis. This algorithm could pop out the vascular structure and facilitate the axis extraction. Considering the analogical character of Gabor filter with receptive field of simple cell, we simulated the feature detective neuron by Gabor filter to obtain the contextual information, and then enhanced vessels under co-circle constraint. The anisotropic diffusion was involved afterward to removed simultaneously enhanced noises.Axes were extracted from enhanced angiograms by Nonmaximum suppression of Gabor response, which was as the result of multi-scales and multi-orientation Gabor filtering. Then we used the thresholding method to remove part of pseudo axes. Based on it, we connected broken points based on the orientation provided by Gabor filtering, and then found connected component as the final axes. The approach in this thesis was implemented on several angiograms in our experiments. The results showed that it could well extract the angiographic axes and achieve satisfactory, smooth and integrated results.In the end, diameters of vessels were estimated from scales obtained by Gabor filtering. We used the scales corresponding to the maximum Gabor response as the diameters of vessels, which provided necessary data for medical therapy and 3-D reconstruction of vessels.