An adaptive and statistical approach for blood vessel segmentation using genetic algorithm

The detection of linear tree structures such as blood vessels is an important procedure for computer-aided diagnosis and further follow-up in certain diseases. This type of pattern is very common in biomedical images and significant in the computation of blood flow. This paper presents an adaptive and statistical approach for the blood vessel-like patterns enhancement in a noisy environment. The statistical nature of various blood vessel-like patterns is different due to their tree-like geometrical structure, such images can be registered using the related feature. Frangi line enhancement technique is efficient and popular for vessel segmentation. However, to get optimal results the filter parameters are to be adjusted manually. In our approach, automatic selection of Frangi scale ratio is proposed using statistical information for the image structure. The segmentation between vessel and background is performed using adaptive thresholding and optimization. In diagnosing and surgery planning, knowing the exact location of the blood vessel is of great importance. Our method performs better in terms of SNR and image entropy; hence results are very encouraging in a noisy environment.