A glaucoma detection method based on radon domain feature representation

Abstract

The invention discloses a glaucoma detection method based on Radon domain feature representation, which comprises the following steps: step 1: converting the color fundus image into a grayscale image, and performing grayscale image preprocessing; step 2: under n projection angles Radon transform is used to project the preprocessed grayscale image into the Radon domain to obtain a one-dimensional discrete signal; Step 3: Unify the dimensions of the one-dimensional discrete signal, and use biorthogonal wavelet to decompose the one-dimensional discrete signal to extract the approximate coefficient and Detail coefficient; Step 4: Combine the approximation coefficient and detail coefficient of each group of one-dimensional discrete signals into a sample feature input classification detection model to obtain glaucoma detection results; the input parameters for classification detection model training are glaucoma fundus image samples and normal fundus images The samples constitute the eigenvector matrix N and the sample label vector for each row in the eigenvector matrix N. The present invention can accurately detect whether it is glaucoma through the above method.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a glaucoma detection method based on Radon domain feature representation. Background technique [0002] Glaucoma is one of the most common and irreversible blinding diseases in the world. It is mainly characterized by elevated intraocular pressure, optic atrophy and visual field defect. If it is not treated in time, the visual field will be gradually lost until blindness. At present, there is no effective cure for glaucoma, and the only way to slow down the progress of glaucoma is to achieve the purpose of protecting eyesight. However, because the occurrence of glaucoma is insidious and gradual, especially for primary open-angle glaucoma, once vision loss is found and the doctor seeks treatment, it is often in the late stage, and the visual field defect is serious and irreversible. Therefore, early detection and timely treatment of glaucoma are emphas...

AI Technical Summary

Problems solved by technology

However, it is time-consuming and labor-intensive to make a diagnosis by relying on the way of manual film reading by doctors. In addition to the differences in personal experience, the consistency rate of diagnosis for the same patient by different doctors is less than 80%. On the other hand, due to the complexity of glaucoma and individual In the face of the continuous growth of glaucoma patients in my country, this will greatly exceed the burden of doctors, which will inevitably lead to misdiagnosis and missed diagnosis.

On the one hand, this type of method requires familiarity with the pathological changes of glaucoma in order to add prior information to make the learning model more robust when extracting difference features; Affected by lesions, optic cup segmentation has not yet been well resolved due to its fuzzy borders

[0006] Therefore, based on the existing computer-aided diagnosis glaucoma detection technology that needs to add prior information when extracting differential features and heavily relies on the segmentation accuracy of the fundus structure, resulting in reduced detection reliability, a feature representation based on Radon domain is proposed. glaucoma detection method

Images