Video Super-Resolution Of Cataract Phacoemulsification Surgery Video Based On Convolution Neural Network

Ophthalmology-related statistics illustrate that cataract is the leading cause of blinding eye disease.As the post-war baby boomer generation and the world ages,so does the incidence and blindness of cataracts.Surgery is the only way to treat cataracts.Phacoemulsification cataract extraction is a new type of cataract treatment.The cloudy lens in the patient’s eyeball is removed by phacoemulsification and suction.Medical research has proved that phacoemulsification cataract extraction is more effective in cataract extraction than traditional surgical methods,and the postoperative recovery speed is also faster.The rise of deep learning in the field of image processing has spawned many image and video super-resolution reconstruction methods based on convolutional neural networks.Video super-resolution methods is originates from image super-resolution,video consists of sequences of video frames,and video super-resolution methods usually take advantage of inter-frame information for recovery.At present,many high-performance image and video super-resolution methods are carried out in laboratory hypothetical environments,and there is very little work on super-resolution reconstruction of real images and videos.Therefore,video super-resolution methods applied to phacoemulsification surgery video is groundbreaking.The recovered high-definition surgical video will be a good guide for doctors and patients to review the surgical process,and will also be of great benefit to potential related medical research.The specific research content of this paper is arranged as follow:(1)We plan to construct a video super-resolution dataset,which is based on phacoemulsification cataract extraction surgery videos.Firstly,the content presented by cataract phacoemulsification surgery videos is analyzed in detail,and a super-resolution dataset for phacoemulsification surgery video is constructed by combining the characteristics of popular video super-resolution datasets and medical image datasets.Then,on the basis of this dataset,we carried out a series of detailed benchmark tests and comparative experiments.(2)We propose a super-resolution method applied to phacoemulsification cataract surgery videos.The video super-resolution(VSR)process is modularized and divided into three parts: feature extraction,feature fusion and refinement and reconstruction.The specific process is as follow: First,according to the characteristics of phacoemulsification cataract surgery videos,a specific and feasible model is constructed in modules and levels.Second,a series of contrast and ablation experiments are performed on various public benchmark datasets to verify the effectiveness of the proposed VSR method.Finally,to verify the generality of the proposed model,we compare with other state-of-the-art image/video superresolution methods on various image super-resolution test sets.