Image recognition method for fungal keratitis based on rx anomaly detection and texture analysis

Abstract

The invention discloses a fungal keratitis image recognition method based on RX abnormality detection and texture analysis, comprising: acquiring normal corneal nerve images and mycelium images containing only hyphae as training samples; acquiring fundus images of patients with fungal keratitis As a test sample; preprocessing, feature extraction and feature fusion are performed on the normal corneal nerve image in the training sample to obtain the neural features after training sample fusion; Extraction and feature fusion to obtain the mycelium features after training sample fusion; preprocessing, feature extraction and feature fusion to the image in the test sample to obtain the neural features after the test sample fusion and the mycelial features after the test sample fusion; Identify nerves and hyphae in test samples.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to an image recognition method for fungal keratitis based on RX abnormality detection and texture analysis. Background technique [0002] Fungal keratitis is an infectious corneal disease with a high blinding rate caused by pathogenic fungi. Due to its rapid development, it is very easy to cause serious consequences such as corneal perforation, hypopyon, endophthalmitis, etc., so it is extremely harmful , the early diagnosis and treatment of this disease is particularly important. Confocal microscope is a new type of non-invasive corneal imaging examination instrument, which can scan and image the cornea from the four-dimensional (three-dimensional space and time) level in vivo, and provide high-definition and magnification images of all layers of the cornea , allowing people to directly observe and study the pathophysiology of living corneas at the cellular level. [000...

AI Technical Summary

Problems solved by technology

A large number of medical images have caused two problems: On the one hand, reading the examination images of patients has become a very heavy task for doctors. Due to long-term interpretation of images, doctors are prone to fatigue and distraction, resulting in a decline in the accuracy of diagnosis; On the other hand, it is difficult to guarantee that there will be no missed diagnosis and misdiagnosis by relying on the doctor's own experience, and it is difficult to conduct a consistent quantitative analysis of the imaging data, and the quantitative analysis of medical images is an inevitable requirement for the development of medical imaging

Traditional image preprocessing methods such as brightness transformation, spatial filtering, and histogram processing do not perform very well, and cannot effectively remove background interference.

Images