Deep convolutional neural network-based optic disc positioning method

Abstract

The invention discloses a deep convolutional neural network-based optic disc positioning method, which comprises the following steps of: processing an eye fundus image to obtain a corresponding saliency map, and positioning an optic disc saliency region on the basis of the saliency map according to brightness; constructing a deep convolutional neural network, adopting a sliding window to select anarea block completely including the optic disc and completely not including the optic disc on the fundus image to form a training and testing image set, and training the deep convolutional neural network to obtain a classification model; and sending the obtained optic disc salient region graph into a classification model until the sent salient region graph is determined to be the optic disc, otherwise, continuing to search the optic disc salient region until the optic disc salient region is determined to be the optic disc. Hierarchical automatic positioning of the visual disc is realized, andthe positioning accuracy is high.

Description

technical field [0001] The invention relates to the field of medical image processing, in particular to an optic disc positioning method based on a deep convolutional neural network. Background technique [0002] The positioning and detection of the optic disc is a key step in the analysis and processing of fundus images. On the one hand, as an important feature of retinal fundus images, the optic disc can be used as an important reference for subsequent fundus image processing. For example, according to the relative position of the optic disc and the macula, it can be Distinguish between left and right eyes in fundus images. The optic disc is the entrance of the blood vessels into the retina, and the positioning of the optic disc also has a great effect on the tracking of retinal blood vessels. On the other hand, in addition to physiological information, optic disc localization and detection also play a crucial role in computer-aided diagnosis. Because when some eye disea...

AI Technical Summary

Problems solved by technology

This method has high robustness on different images, but this type of method is complex, requires a lot of calculations, and is highly dependent on blood vessel segmentation. When the blood vessels in the fundus image are not obvious, this method is less effective. Poor, so not suitable for large-scale image screening and research

Images