Method and system for recognizing lattice degeneration and holes in wide area fundus images based on deep learning

Abstract

The invention relates to a method and system for recognizing lattice degeneration and holes in wide area fundus images based on deep learning. The method comprises the steps that the wide area fundusimages are input into a convolutional neural network, and whether peripheral retina lattice degeneration or holes exist in the wide area fundus images or not is judged; and when it is judged that theperipheral retina lattice degeneration or the holes exist in the wide area fundus images, a saliency map is adopted to position the lattice degeneration position or hole position in the wide area fundus images. The method and the system can assist an ophthalmologist in reading the wide area fundus images of patients more accurately, more conveniently and more quickly.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, and more specifically, to a method and system for identifying grid change holes in wide-area fundus images based on deep learning. Background technique [0002] The disadvantages of existing inspection techniques for peripheral retinal lattice degeneration and holes are high requirements for inspectors, time-consuming and labor-intensive inspections, and the possibility of implementing large-scale crowd screening for peripheral retinal lattice degeneration or holes is low. The main reasons are as follows: [0003] 1. Screening for peripheral retinal lattice degeneration or holes requires high technical requirements for ophthalmologists and requires years of clinical experience in fundus diseases. Usually, community hospitals or low-level general hospitals and physical examination centers do not have professional ophthalmologists, so they cannot conduct comprehensive dete...

AI Technical Summary

Problems solved by technology

Usually community hospitals or low-level general hospitals and physical examination centers do not have professional ophthalmologists, so they cannot conduct comprehensive detection and evaluation of such lesions

[0004] 2. The amount of information contained in the wide-area fundus image is about 4 times that of the traditional fundus image. The peripheral retina has a variety of grid-like degeneration and holes, and the early features are not obvious. Accurate interpretation of this fundus image requires ophthalmologists to undergo professional training and comparison. Accumulation of long-term experience

[0005] 3. Manual inspection of the peripheral retina requires mydriasis and the close cooperation between the fundus doctor and the patient. Most elderly patients or children do not cooperate well. The inspection takes a long time, and it is easy to miss the diagnosis of the disease due to insufficient cooperation.

[0006] In summary, due to the limitations of ophthalmologists' examination techniques in primary hospitals, and the time-consuming detailed inspection of peripheral retinal lattice degeneration and holes, it is difficult to carry out screening of peripheral retinal lattice degeneration and holes in large-scale population, which can lead to the failure of most patients. Delay in diagnosis and treatment, some patients have retinal detachment due to the progress of the disease, resulting in visual damage, and the visual damage caused by the disease is irreversible, which has caused great losses to individuals, families and society

Images