Faster RCNN-based papillary thyroid carcinoma ultrasonic image recognition method and system

Abstract

The invention discloses a Faster RCNN-based papillary thyroid carcinoma ultrasonic image recognition method and system. The method comprises the steps that a training sample including an ultrasonic image and the corresponding diagnosis result is acquired, wherein the ultrasonic image is the image obtained by performing ultrasonic imaging on the papillary thyroid carcinoma affected region of the patient; a Faster RCNN network to be trained based on a ZF network is trained by using the training sample so as to obtain a corresponding trained model, wherein the Faster RCNN network to be trained is the Faster RCNN network obtained after connection of the fourth layer and the fifth layer in the shared convolutional layer; and when the ultrasonic image to be detected is acquired, the ultrasonic image to be detected is inputted to the trained model so as to obtain the detection result outputted by the trained model. The used Faster RCNN network is obtained after connection of the fourth layer and the fifth layer in the shared convolutional layer so that the effect that the output characteristics of the fourth layer and the fifth layer can be converged into a path of characteristic flow can be achieved, and the accuracy of image recognition can be effectively enhanced.

Description

technical field [0001] The present invention relates to the field of image recognition, in particular to a method and system for ultrasonic image recognition of papillary thyroid carcinoma based on Faster RCNN. Background technique [0002] Image recognition is a kind of research hotspot. With the development of digital image technology and the needs of practical applications, a new difficulty has emerged, that is, the output of the result is not required to be a complete image, but the processed image is then segmented and described to extract effective features. Then judge and classify, this is a new technological science developed in the past 20 years - image recognition. Its main content is to study the classification and description of certain objects or processes, and to develop a machine vision system that can automatically process certain information to replace the traditional tasks of manual classification and identification. However, there are relatively few appl...

AI Technical Summary

Problems solved by technology

Once the features change (such as distortion, flipping, lighting changes, damage, etc.), the effect of these algorithms will become worse, and its applicability is not strong

Images