Computer Aided Detection Of Brain Lesions

With the continuous improvement of technology,magnetic resonance imaging(MRI)has become an important means for detecting various tissue lesions in human body due to its high-quality imaging effect,and it is the most widely used in brain lesion detection.Accurate segmentation of MRI brain tumors provides reliability for rapid clinical diagnosis and treatment.However,the diversity of brain tumor cases has brought difficulties to the study of brain tumor segmentation.Based on the existing brain tumor image segmentation algorithm,this paper improves and optimizes the existing algorithms by using potential energy theory,region growing algorithm,particle swarm algorithm and FCM algorithm,and better realizes the accurate segmentation of brain tumor images.First of all,this paper briefly introduces the research status of brain tumor image segmentation at home and abroad,the commonly used brain tumor segmentation algorithm and MR imaging principle.Provide a theoretical basis for subsequent research.Secondly,this paper proposes an improved region growing algorithm.First,by using the concept of "potential energy" in electronics,the pixel value of each point of the image is regarded as the amount of charge of the point charge,and the size of the "charge amount" of each pixel is used for coarse segmentation.Then,the quaternion is introduced in the region growth.According to the quaternion representation,the pixel values of the seed point are represented by the pixel values of the seed point,and the quaternion vector product is combined with the particle swarm algorithm to determine the growth rule.Using the improved region growing algorithm described above,the segmentation image is segmented accurately,and the segmentation of brain tumor images is achieved.Thirdly,based on the optimized particle swarm optimization algorithm and fuzzy clustering,an improved FCM algorithm is proposed.The clustering center is obtained by the optimized particle swarm optimization algorithm to solve the problem of poor selection of clustering centers.The penalty term of spatial information is added to the FCM algorithm,which solves the problem that the traditional FCM algorithm will not fully image segmentation,resulting in pixel points.The problem of incorrect classification is to achieve segmentation of brain tumor images.Finally,the two segmentation methods proposed in this paper are compared and compared,and the conclusions that each has its own advantages are obtained.The thirdchapter uses less time but has lower precision,while the fourth chapter has higher precision but longer time.