Knee joint disease ultrasonic diagnosis method based on deep learning multiple channels and graph embedding method

Abstract

The invention discloses a knee joint disease ultrasonic diagnosis method based on deep learning multiple channels and a graph embedding method, and the method comprises the following steps: carrying out the preprocessing of an effusion region in a knee joint ultrasonic image through employing an snake algorithm, and then inputting the effusion region into a defined network model for semantic segmentation; on the basis of a Resnet network structure, training a knee joint ultrasonic image in a data set by utilizing a graph embedding method of secondary training, and finally performing verification by utilizing tests of a segmentation network and a classification network. According to the invention, the knee joint ultrasonic image is segmented and trained by using the thinking of multi-channel superposition and a graph embedding method; disease categories can be distinguished according to whether hydrops areas in different knee joint disease ultrasonic images are accompanied by the difference of synovial membrane thickening or not, the situation that knee joint ultrasonic image judgment completely depends on naked eyes and personal judgment of doctors is avoided, the problems of subjectivity and personal errors are eliminated, and the whole segmentation and classification recognition method is simple, efficient and accurate in diagnosis.

Description

technical field [0001] The invention relates to the technical field of machine vision, in particular to an ultrasonic diagnosis method for knee joint disorders based on deep learning multi-channel and image embedding method. Background technique [0002] The knee joint is the most complex joint in the human body, and people of all ages are susceptible to infection or injury. Common knee-related diseases include synovitis, synovial thickening, and cysts. Medical images are a common and important method for knee joint diagnosis at present. The liquid area of ​​the lesion appears as a darker black area in the image. Doctors use this area as the main basis for judgment. At the same time, the accuracy of the area demarcation also affects The doctor's correct diagnosis. At present, the diagnosis of common diseases of the knee completely depends on the doctor's naked eyes and personal judgment, which wastes a lot of manpower and material resources, and has certain subjectivity and...

AI Technical Summary

Problems solved by technology

At present, most of the processing of knee ultrasound images still stays on the traditional algorithm. The deep learning algorithm, which is widely used in image processing in various fields, has not been effectively applied to ultrasound images, so its application and knee ultrasound Images are necessary and have important application context

Images