Thyroid nodule automatic grading method and system

Abstract

The invention provides a thyroid nodule automatic grading method and system. The thyroid nodule automatic grading method comprises the following steps: processing a thyroid nodule ultrasonic image to be classified by using a pre-trained total classification network model and outputting a classification result; during training, manually marking and dividing each thyroid nodule ultrasonic image manually into a training set, a verification set and a test set; extracting a shallow feature map; extracting a deep feature map based on an evaluation index; processing the deep feature map and outputting a fine classification result based on each evaluation index; fusing five groups of deep feature maps output by the second feature extraction model; processing the fused deep feature map, and outputting a TI-RADS grading result; and constructing a loss function, and training a total classification network model based on a manual annotation result. Tediousness of manual grading is avoided, diagnosis efficiency and accuracy are improved, and doctors can be assisted in diagnosis in clinical medicine.

Description

technical field [0001] The present invention relates to the technical field of ultrasonic image processing, in particular to an automatic grading method and system based on thyroid nodules. Background technique [0002] Thyroid nodule is a common thyroid disease, in which the incidence of malignant thyroid cancer is about 5% to 15%. Ultrasonography is an effective method for diagnosing thyroid nodules. However, due to the complexity and overlap of thyroid nodule images obtained by ultrasound, it is difficult to accurately identify some nodules with inconspicuous ultrasound features, resulting in the detection of such nodules. Moreover, the ultrasonic diagnostic standards for thyroid nodules are not uniform, so that the accuracy of ultrasonic diagnosis is affected by subjective factors such as ultrasonic equipment and doctor experience, which makes it difficult to qualitatively examine the results. Based on this, the American College of Radiology (ACR) proposed the TI-RADS (...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the above problems, the present invention provides an automatic grading method and system for thyroid nodules, which aims to solve the cumbersome and inaccurate manual classification of thyroid nodules in the prior art. technical problem

Images