Study On Corneal Nerve Morphology Assessment Based On In Vivo Confocal Microscopy Image

The geometrical changes of corneal nerves are an important basis for early evaluation of corneal diseases and neurological diseases.In vivo confocal microscopy(IVCM)technology can be used to observe the corneal nerve in vivo.However,due to the small imaging range of IVCM and only single shots,the current clinical evaluation of corneal nerve morphological changes only stays at the local calculation stage,therefore this method has greater subjectivity and limitations.This study uses computer-aided technology to realize the automatic stitching of multiple corneal images of the same patient and the calculation and analysis of corneal nerve morphology.The specific content is as follows:(1)Proposed an automatic stitching method for IVCM images.Manual stitching of multiple corneal images is not only time-consuming and labor-intensive,but may also cause subjective errors.Therefore,this paper proposes a corneal image automatic stitching method based on feature point matching,and compares and analyzes the SIFT operator and the feature point matching result based on convolutional neural network.At the same time,the stitching module in the open source library Open CV is used to realize the corneal image.Finally,the stitching results of corneal nerves of different images under the action of spatial transformation are investigated.(2)A global corneal nerve labeling method is proposed.Due to the different imaging conditions of a single IVCM,the stitched corneal images have obvious regional boundaries.Therefore,this research proposes an image brightness consistency algorithm.In order to extract effective corneal nerve bundles,this paper first enhances the neural boundary by Gaussian high-pass filter,and then uses boundary tracking algorithm to complete the global neural labeling and grouping,and construct the neural tracking map.(3)The neuromorphological measurement indexes of corneal stitching images and single images are compared.This article first analyzes the measurement standards of nerve tortuosity,and proposes a method for evaluation of tortuosity based on multi-scale parametric spline curve fitting.After that,the qualitative indexes of the nerve density of the IVCM images after 4 sets of stitching and the nerve density of a single IVCM image under random sampling were compared.The experimental results prove that there are obvious differences between the two.This article collects the binocular IVCM images of 16 patients with mild dry eye,and continuously collects 300 to 600 images in a single eye.At the same time,in order to promote the development of IVCM image processing technology,a corneal confocal microscopy image database(CCMID)was created to collect the IVCM image data of each patient for reference by clinicians and researchers.The database can also be used for ophthalmology medical teaching and related product research and development.It is particularly critical for the implementation of precision medical imaging,"medical imaging + artificial intelligence" and other interdisciplinary research,and it is expected to further improve diagnostic efficiency and objectivity.