High-specificity diabetic retinopathy characteristic detection method and storage equipment

Abstract

The invention relates to the technical field of medical image processing, in particular to a high-specificity diabetic retinopathy characteristic detection method and storage equipment. The high-specificity diabetic retinopathy characteristic detection method comprises the following steps: carrying out focus characteristic detection on a lesion area of a fundus image through a preset step; preprocessing the fundus image processed by the preset step; extracting a main blood vessel of the preprocessed fundus image; performing optic disc delineation and macular foveal delineation on the preprocessed fundus image according to the main blood vessel; and further perfecting the specificity of focus feature detection according to the optic disc delineation result, the macular fovea delineation result and the main blood vessel. Compared with a method which can only reflect the eye fundus image level, an eye fundus image classification method based on the image level or a focus characteristic extraction method only through deep learning can reduce focus detection errors by directly obtaining the positions, types and numbers of red and bright color focuses, and the focus feature detection specificity is improved.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence and medical image processing, in particular to a highly specific diabetic retinopathy feature detection method and a storage device. Background technique [0002] Diabetic retinopathy (Diabetic Retinopathy, DR) is a chronic and imperceptible disease, which is one of the main causes of human blindness. Therefore, in the normal population, early and regular screening for DR is very necessary. The screening of large-scale fundus images increases the burden on doctors. In order to reduce the workload of doctors and improve efficiency, it is of great significance to realize the automatic detection of lesions in the DR screening system. [0003] Detection of bleeding points, hemangiomas (red lesions) and exudates (bright lesions) are particularly important in studies of early detection of DR. However, the similarity of anatomical structures and lesions in fundus images in some features...

AI Technical Summary

Problems solved by technology

However, the similarity of anatomical structures and lesions in fundus images in some features and the quality of fundus images caused by imaging hardware conditions make it difficult to detect DR.

[0004] Throughout 2017, many automatic fundus image analysis systems for artificial intelligence DR screening at home and abroad have mostly pursued high sensitivity, high accuracy or high accuracy, while ignoring medical specificity, resulting in many unnecessary errors. Not only will it not be of much help to the majority of doctors, but the specificity is extremely low! The artificial intelligence system was originally designed to better assist doctors in the interpretation and grading of DR lesion characteristics. The misjudgment rate is extremely high, which not only seriously affects DR grading, but also forces doctors to spend more time excluding lesions that may be misjudged Or characteristic points, these misjudged lesions are more helpful to doctors! If the specificity is too low, it is difficult to put it into practical application!

Images