Study On Parallel Signed Distance Transformation And It's Application In Brain Extraction

The process of separating brain tissue from skull,eyeball,skin,fat and other tissues in cerebral MRI images is called brain tissue extraction.It is an important processing step of brain MRI image analysis and has important applications in fMRI image registration,brain tissue segmentation,brain volume measurement,etc.Fast and accurate extraction of brain tissue has important applications in clinical and research.This paper proposes a parallel brain tissue extraction method on CUDA parallel computing platform.This method combines parallel BET algorithm and parallel level set method,and is a hybrid algorithm.The main contributions of this paper are as follows:1.In order to further improve the calculation speed of parallel level sets,this paper studies the parallel calculation of signed distance function,which plays an important role in the evolution of level sets,and proposes a signed distance function calculation method based on normal radiation on the basis of source point scanning method.The algorithm firstly obtains the signed distance of pixels in the normal direction of the closed curve in the image by using the normal emission method,and then quickly calculates the signed distance function according to the source point scanning algorithm,thus reducing the times of scanning all the pixels in the image and improving the calculation efficiency.The parallel algorithm is also implemented on CUDA platform.Experiments show that the parallel algorithm has better computational efficiency than the fast dimensional reduction method under the same computational accuracy.2.In order to achieve fast and accurate brain tissue extraction,a new hybrid parallel brain tissue extraction method is proposed by combining parallel BET algorithm and parallel level set algorithm.The method firstly uses parallel BET algorithm to extract brain tissue from the input brain MRI image to obtain a preliminary result,and then uses the result as the initial contour of the level set to carry out level set evolution to obtain a more accurate result.During the evolution of the parallel level set,the parallel signed distance function calculation method proposed in this paper is adopted to initialize and reinitialize the level set function,which greatly improves the calculation time.Brain tissue extraction can be completed in only 1.5 seconds,thus realizing real-time processing.3.In order to improve the extraction accuracy,a user interaction system is designed in this paper.Users can easily modify the parameters in BET and level set through the keyboard until satisfactory results are obtained.Because the single processing time is very fast,the user can still obtain accurate brain tissue extraction results in a very short time despite user interaction.