Macular central fovea detection method and device, computer equipment and a storage medium

Abstract

The invention relates to a macular central fovea detection method and device, computer equipment and a storage medium. The method comprises the steps: inputting a to-be-processed image into a segmentation network, and carrying out the feature extraction of the to-be-processed image through employing all first convolution modules at all levels, thus obtaining an initial probability graph and an intermediate feature graph output by the last first convolution module on each level; performing feature fusion and spatial coding on the intermediate feature map output by each level by using a multi-level feature fusion spatial coding network to obtain a final position feature map of the macular fovea; multiplying the final position feature map of the macular fovea with the initial probability mappixel by pixel to obtain a final prediction map; and finally, determining the position of the macular central fovea according to the final prediction graph. The problem that the accuracy of the targetdetection model for detecting the macular area is affected by unclear edges of the macular area, low contrast and other factors is avoided, and accurate detection of the macular central fovea is achieved.

Description

Technical field[0001]The present application relates to a computer technology, and in particular, a method, apparatus, computer device, and storage medium relating to a macular center concave.Background technique[0002]The macrotic area is located in the middle of the polarity of the pole, which is rich in photochane epithelial cells, and is an important physiological structure in the human eye. There is a small depression of the macular center, called the macular center concave, is the most sensitive part of the visual visual on the retina. In the fundus image, the macular center concave positioning is of great significance to many retinal lesions (such as pathological myopia, diabetes retinopathy, etc.).[0003]In the relevant technique, the model training is first carried out by a training image of 10,000 tabs, and the target detection model is obtained; then, the macular region is detected directly through the target detection model.[0004]However, in the related art, the edge of th...

AI Technical Summary

Problems solved by technology

[0004] However, in related technologies, factors such as unclear edges of the macular area and low contrast will affect the accuracy of the target detection model in detecting the macular area.

Images