Fundus Optic Disc Location And Segmentation Based On Adaptive Fractional Darwinian Particle Swarm Optimization Algorithm

Image segmentation technology is an important research direction in the field of computer vision.Medical image segmentation as a kind of image segmentation,has also become one of the research directions of researchers.Nowadays,with the widespread popularity of computer technology,the diagnosis methods of doctors in individual fields are inseparable from medical imaging technology.The computer can use image segmentation technology to mark the characteristic areas so that the doctor can perform pathological analysis based on the marked characteristic areas and propose corresponding diagnosis and treatment plans.This technique is often used in the diagnosis and analysis of skin,bone,organ tissue,fundus and other fields.The segmentation of the fundus optic disc can effectively assist ophthalmologists in diagnosing glaucoma and other ophthalmic diseases.The use of traditional segmentation algorithms to segment the fundus optic disc area has disadvantages such as susceptibility to other tissues,slow segmentation speed,and low segmentation accuracy.In view of the shortcomings of these traditional algorithms,this paper improves and optimizes them.The main work of this thesis includes the following processes:(1)The image preprocessing stage.In the image preprocessing stage,in order to effectively extract the optic disc area of the fundus,the accuracy of the algorithm should be improved by removing the interfering factors such as the blood vessel area of the fundus image.The color interval conversion of the fundus image is used to improve the accuracy of the algorithm.The gray component performs morphological closing operation and median filtering operation,which can finally effectively eliminate blood vessels in the fundus image.The high-brightness optic disc area is well preserved in the gray image to provide a guarantee for post-processing.(2)In the image segmentation stage,the traditional particle swarm optimization algorithm is used to segment the fundus optic disc region.Although this segmentation method can effectively segment the optic disc region,the segmentation accuracy is relatively low,and it is easy to fall into the local maximum.Aiming at the defects of the traditional algorithm,this thesis makes optimization and improvement.Firstly,the improved particle swarm optimization algorithm is improved to Darwin algorithm,and then the fractional order optimization is introduced on the basis of Darwin particle grouping optimization algorithm.It is best to introduce adaptive control in the fractional level.This operation can adjust the fractional order according to the state of the particles.By comparing SSIM and PSNR,several optimization algorithms are compared.The SSIM index values of traditional particle swarm optimization algorithm,Darwinian particle swarm optimization algorithm,fractional Darwinian particle swarm optimization algorithm and Adaptive Fractional Darwinian particle swarm optimization algorithm are 0.7548,0.7655,0.7672 and 0.7730 respectively,and the peak signal-to-noise ratio is 24.29db,24.96db,25.33db and 26.22db respectively.The analysis shows that the introduction of Adaptive Fractional Darwin particle swarm optimization algorithm can improve the performance and accuracy of the algorithm.(3)In the image edge detection and ellipse fitting stage,the Canny operator is used to detect the segmented real disc contour.Since the optic disc contour is similar to the elliptical contour,the least squares ellipse fitting is used to fit the edge contour,and finally the fitting results are displayed in the original fundus image to complete the final optic disc positioning.The algorithm of this thesis is compared and analyzed on the gold standard of the ORIGA and DRION-DB data sets.The experimental results show that compared with the traditional threshold segmentation algorithm and the preliminary improved particle swarm optimization combined with the threshold segmentation algorithm,the algorithm of this paper the gold standard of manual segmentation with experts in the data set has a higher overlap rate,and the segmentation accuracy has been greatly improved.Compared with the traditional particle swarm optimization algorithm,the Adaptive Fractional Darwinian particle swarm optimization algorithm greatly improves the convergence performance,and can be well applied to the fundus image segmentation.