Research On Macular Degeneration Diagnosis Method Based On Deep Learning

With the development of artificial intelligence and deep learning technology,deep learning has been widely studied and applied in the medical field.From the perspective of computer-aided diagnosis,deep learning,especially convolutional neural network technology,has natural advantages in the processing and analysis of medical images.Ophthalmology is a subject that is highly dependent on image examination.This topic focuses on aged-related macular degeneration and provides a technical basis for computer-assisted macular degeneration diagnosis through deep learning techniques and fundus images.Macular degeneration is a blinding eye disease that requires timely diagnosis and treatment and regular review to control the development of the disease.The research objectives of this paper include building an independent macular degeneration fundus image database,researching classification and prediction methods of macular degeneration fundus images based on deep learning technology,training and optimizing diagnostic models,and improving the accuracy of computer-aided macular degeneration diagnosis.The main work of this article is summarized as follows:1.This paper establishes an independent macular degeneration fundus image data set by collecting data.This data set contains 2763 fundus images,which are initially classified by medical professionals,and provide a data basis for the research of macular degeneration diagnosis.2.This article analyzes the quality of the fundus image data set,and proposes a non-reference fundus image quality evaluation method for image problems such as blur,overexposure,and underexposure that affect the quality of the dataset.This method has certain universality in evaluating the fundus image quality,and is conducive to the establishment of high-quality fundus image data sets.3.This paper studies the relevant knowledge of convolutional neural networks in deep learning,uses a suitable transfer learning method to train the macular degeneration dataset,and compares the effects of different convolutional neural network models on the diagnosis of macular degeneration.Compared with the direct training method,it proves the effectiveness of convolutional neural network technology and transfer learning in the diagnosis of macular degeneration.4.Based on comparative experiments,this paper makes innovative optimizations in both network structure and diagnostic model design.A new network structure XVGG and a bilinear network diagnostic model combined with the macular region are proposed.The model achieved 92.4% accuracy on the macular degeneration dataset.