Research On Extracting Keratoconus Contour Based On Machine Learning

Machine learning,as one of the most important techniques in the field of artificial intelligence,has a bright future in ophthalmology.This paper presents a new method for extracting keratoconus contour based on machine learning.By visualizing Corvis ST corneal image data obtained by corneal biomechanical analyzer,the contours of keratoconus can be extracted to further achieve the purpose of disease diagnosis.Different from traditional contour extraction algorithms,the machine learn-based keratoconus contour extraction method proposed in this paper can effectively reduce the contour extraction error.Firstly,DeckSENet neural network is used to segment and process the acquired original data image.Secondly,OTSU algorithm is used to calculate a reasonable threshold to obtain more accurate keratoconus contour results.Finally,a comparative test was conducted on the Corvis ST corneal image dataset.The results show that the method proposed in this paper can be used to extract a more accurate and complete corneal contour,which can preserve the corneal body to a maximum extent and reduce the corneal contour loss to a minimum extent.The work of this thesis consists of the following two points:(1)Design a corneal contour extraction algorithm of Corvis ST image based on OTSU algorithm.Aiming at the problems of insufficient extraction accuracy,severe damage to the cornea body and rough corneal edges of the keratoconus contour extracted by the existing contour extraction methods,the edge detection algorithm was used to replace the contour extraction part of the traditional contour extraction methods.In the key process of image processing,the timely use of mathematical morphology operation is helpful to obtain a more complete corneal contour,reduce unnecessary losses and improve the accuracy of contour.The method proposed in this paper can not only maintain a high accuracy of keratoconus contour but also eliminate the problem of rough corneal contour edge and improve the effectiveness of corneal contour to a certain extent.(2)Based on the unsupervised learning network,DeckSENet is designed.The purpose of this paper is to improve the efficiency and precision of the present method in the extraction of keratoconus profile by using the above methods.Through careful analysis of current keratoconus contour extraction methods and systematic study of neural network,it is found that it has a good performance in the field of image segmentation.The DeckSENet neural network is designed.This paper attempts to design a machine learning-based keratoconus contour extraction algorithm by adding DeckSENet neural network before the above OTSU keratoconus contour extraction algorithm.This algorithm further improves the effectiveness and accuracy of keratoconus contour extraction results of Corvis ST corneal image data set,and can achieve better extraction results.