The Research Of MR Image Segmentation Algorithm In Stroke

With the maturation of medical image processing technology and the improvement of related hardware technology,computer-aided diagnosis has been widely used in clinical disease diagnosis,common pathological analysis and patient treatment,and especially in the study of brain diseases,it is imperative that computer-aided diagnostic techniques help doctors complete the treatment.In most brain diseases diagnosis,nuclear magnetic images are taken as an important diagnostic basis,so the research on brain magnetic image processing technology has become a new research hotspot.Among them,the accurate segmentation of brain tissue and lesions not only helps doctors quickly understand the condition,but also plays an important role in determining the follow-up treatment plan and 3D visualization.Although there are many mature theories and methods in medical image segmentation at present,due to the difficulties encountered in the imaging process,such as device design defects,illumination differences,and brain tissue incompatibility.Therefore,for different diseases,it is necessary to combine with actual needs to produce image processing methods with practical value.In this paper,taking the segmentation of stroke's MR images as the starting point,I analyzes the problems that may be encountered in the segmentation of stroke's MR images,and proposes some improved methods in combination with image processing techniques.The main work includes the following aspects:1.The medical image segmentation methods such as threshold segmentation,fuzzy clustering,random field model,active contour model and so on are studied.Finally,the segmentation is implemented by improved fuzzy C-means(FCM).FCM algorithm introduces the theory of fuzzy sets,which can show the outline of brain tissue more clearly.However,this method has the disadvantages of large amount of calculation and not considering spatial information.To deal with these shortcomings,the Canny operator is first used to eliminate background noise while eliminating the noise,which reduces the computational complexity of subsequent segmentation algorithms.Secondly,the neighborhood information of the center pixel is fully considered,and the weight assignment method based on the gray gradient is used to make a improvement.Experimental results show that compared with the traditional FCM algorithm and FCMS algorithm,the improved FCMS algorithm has greatly improved the clustering effect.2.The method of segmentation based on regional growth in stroke MR images was studied.The traditional regional growth requires manual selection of seed points.This method not only requires the operator to have certain medical knowledge,but also a time-consuming and laborious work.In this paper,the cluster center obtained by the improved FCMS algorithm is used as an alternative seed point,from which to select,and combined with the characteristics of stroke MR images to develop the corresponding regional growth criteria,thus completing the regional growth improvement.The experimental results show that the improved algorithm has better segmentation accuracy,and its segmentation efficiency has been greatly improved compared to the region growth based on manual mode.