Font Size: a A A

Detection Of Brain Tumor Automated Segmentation And DT - MRI In MRI Images In Nervous Navigation

Posted on:2012-03-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y Y LianFull Text:PDF
GTID:1108330434473415Subject:Medical informatics
Abstract/Summary:PDF Full Text Request
Brain Tumor is seriously endangering the health and life of humans, and neurological surgery is a most effective treatment for it. But it is risking and challenging, since needing not only completely removing the lesions but also keeping the important neural structure intact.Therefor, precisely positioning the brain tumor and normal tissue is the key of neurosurgery.Image Guided Neurosurgery System (IGNS) is widely used in neurosurgery, because it can improve surgery accuracy, shorten surgery operation time, reduce surgical trauma and complications. However, navigation accuracy is suffered from the blurred brain tumor boundaries, difficulty of accurately extracting the fiber bundles and other factors when IGNS is used to resect the brain tumor. Therefore, automatic and accurate detection and segmentation of brain tumors, accurate extraction of nerve fibers is the key factor to improve the accuracy of IGNS. In this paper, for the purpose of improving the accuracy of IGNS, the theory and methods of automatic detection and segmentation of brain tumors, accurate extraction of fiber bundles are systematically researched. The main research content of the full text as following three parts:Part1a method of automatic detection and segmentation of brain tumor in MRI image is proposed based on brain symmetry and sliding-window technique. First, automatic detection of the tumor from MRI images is Milled by a Radial basis function Support Vector Machine based on extracting gray-scale and symmetry features. Sliding window technique, which are based on detecting and extracting the most dissimilar region, are used to automated segment tumor. Sliding window techniques provides a new idea and method for brain image segmentation. Seen from the experimental results, the method can provide surgeons with auxiliary diagnosis whether there is a tumor or not and whether the brain tumor is benign or not, segment tumor accurately, and keep consistent with the average segmentation results from three experienced doctors.In part2, compressed sensing is applied in the DT-MRI research. In order to solve the probledms in DT-MRI, such as long-scanning time, large volumes of data and time-consuming post processing, based on the sparsity of DT-MRI data in Fourier space, the Star-ray sampling and conjugate gradient algorithm for reconstructing images is introduced in the DT-MRI image scaning process, which maintains the high accuracy of the images, also reduces sampling frequency, saves scanning time and reduces volumes of data. Experiment results show that the optimized compressed sensing method using DT-MRI can save about30%imaging time while ensuring the image quality.Part3is about an optimized fiber bundles extraction method based on Hamilton-Jacobi equation. By studying the drift of the seed point in the fiber bundles extraction, it is found that to some extent the uncertainty of the seed point location affects the result of the fiber bundles extraction. In order to maintain the stability of extracted fiber bundles, in an innovative way contour tracking algorithm is used to automated selecting the seed point for the fiber bundles extraction, which improves the automation and stability of the fiber bundles extraction.Improved the original method of selecting two seed points, and select a seed point extraction of fiber bundles, reducing the computational comsumption and improving the stability. Through etermining a spherical region with seed point as center and a certain radius, the scope of subsequent calculations is narrowed, which greatly reduces the computation amount. Experimental results show that the method improves the efficiency of fiber bundles tracking, and also increases the degree of automation.In summary, the method of automatic detection and segmentation of brain tumor, and the optimized fiber bundles extraction method with automatical seed point selection provide possibility for highly precision and intelligent IGNS; while the use of Compressed Sensing in DT-MRI imaging reducing imaging-scanning time makes possibe wide applications of intra-surgery-MRI IGNS.
Keywords/Search Tags:Brain Tumor Detection, Brain Tumor Segmentation, Fiber BundlesExtraction, Compressed Sensing, DT-MRI, Image Guided Neurosurgery, MRI
PDF Full Text Request
Related items