An automatic segmentation method for retinal cystic edema

Abstract

The invention discloses an automatic segmentation method for retina saccular oedema. The method comprises a pretreatment step of removing speckle noise of an OCT image, using a three-dimensional diagram search method to divide a retina into eleven layers, and excluding macular holes and blood vessel shadows to obtain areas of interest of the retina in the first layer to the seventh layer; a rough cut step of extracting 23 texture features in the areas of the interest, adopting the principal component analysis method to perform feature selection, and using an Adaboost classifier for training to obtain a rough cut result; and an exact cut step of using a mathematical morphology algorithm to obtain a background point and a foreground point required by a graph cut algorithm, training parameters of the graph cut algorithm, and obtaining a exact cut result using the fully-automatic graph cut algorithm. The method provided by the invention can still achieve exact cut of the saccular oedema when two illnesses of the macular hole and the saccular oedema exist at the same time.

Description

technical field [0001] The invention relates to an automatic segmentation method for retinal cystic edema, belonging to the technical field of medical image processing and analysis. Background technique [0002] The macula, located in the center of the retina, is the area of ​​sharpest vision. Macular disease can severely affect central vision. A variety of diseases can cause cystoid macular edema, such as diabetic retinopathy, age-related macular degeneration, retinal vein occlusion, retinal inflammation, etc. Another macular disease--macular hole, which refers to the tissue defect from the inner limiting membrane of the macular retina to the photoreceptor cell layer, seriously affects the central vision, and mostly occurs in the elderly over 60 years old. Both of the above-mentioned maculopathy may lead to a sharp decline in central vision and even blindness. We urgently need a method that can measure the volume of cystoid edema when multiple retinal diseases coexist, so...

AI Technical Summary

Problems solved by technology

[0003] So far, many researchers have devoted themselves to the segmentation of cystoid macular edema. Among them, semi-automatic methods, such as Amir H. Kashani et al., manually mark cystoid edema and measure the volume with the help of software. Yalin Zheng et al. Use the mouse to click on the position of the cystic edema, and then get the outline of the cystic edema according to the algorithm. The cystic edema is many and small. These semi-automatic methods are not only time-consuming, but also a lot of work.

Of course, many researchers have proposed fully automatic methods. For example, Gary R. Wilkins et al. used bilateral filtering to denoise, selected slices with cystic edema in the 3D data, and used threshold segmentation to obtain cystic edema. These Algorithms can achieve the segmentation of cystoid edema, but these algorithms can no longer be used when a macular hole is present

Images