Eyeball motion segmentation positioning method based on cyclic residual convolutional neural network

Abstract

The invention discloses an eyeball motion segmentation positioning method based on a cyclic residual convolutional neural network. Acquiring an eye position picture of a testee, inputting the eye position picture into the first-stage cyclic residual convolutional neural network model, and processing the eye position picture to obtain a rotated eye position picture; inputting the first-stage cyclic residual convolutional neural network model again to obtain a binary mask, cutting to obtain left and right eye detection area pictures, and inputting a second-stage cyclic residual convolutional neural network model to detect to obtain eyelid and cornea masks of the two eyes; obtaining a proportional scale according to the circular mark on the forehead; the eye position image is processed by the eyelid cornea mask to measure the eyeball movement, the pixel distance of the six extraocular muscle functions is obtained, and the actual size value is obtained through conversion. According to the method, the technical performance is stable, the segmentation accuracy is high, eyeball movement can be rapidly and automatically detected, errors caused by manual measurement are avoided in a computer-assisted image processing mode, and therefore the accuracy and objectivity of eye muscle disease diagnosis are improved. The invention also provides possibility for physical examination screening, telemedicine and the like.

Description

technical field [0001] The invention relates to an eye image processing method, in particular to an eye movement segmentation and positioning method based on a cyclic residual convolutional neural network. Background technique [0002] Clinically, accurate assessment of eye movements is very important for the diagnosis of ocular muscle-related diseases, especially noncommon strabismus. The movement of the extraocular muscles is very complex, and the extraocular muscles play a synergistic or antagonistic role. For the convenience of diagnosis, the eye position that can display the main action of an extraocular muscle without combining its secondary actions is specified, which is called the diagnostic eye position. The six diagnostic eye positions correspond to the six extraocular muscles (two horizontal rectus muscles + two vertical rectus muscles + two oblique muscles). In clinical practice, doctors usually use qualitative methods to subjectively judge the degree of hypera...

AI Technical Summary

Problems solved by technology

However, this technology cannot avoid image matching errors, which causes inconvenience in clinical application.

Images