Thyroid nodule classification method based on multi-scale feature fusion

Abstract

The invention relates to a thyroid nodule classification method based on multi-scale feature fusion, and the method is characterized in that the method comprises the steps: 1) obtaining an original thyroid ultrasonic image data set, and processing each ultrasonic image; 2) cleaning the original ultrasonic image data set, and removing images which do not meet requirements to obtain a data set containing 2000 high-quality thyroid nodule ultrasonic images; 3) constructing a thyroid nodule ultrasonic image classification network based on a residual network; 4) replacing the residual module with amulti-scale fusion module; 5), based on a residual network, adding a high-resolution channel; and 6), analyzing a network model classification effect based on multi-scale feature fusion and the high-resolution channel. According to the multi-scale feature fusion and high-resolution channel-based network model classification method, the design is scientific and reasonable, a multi-scale feature andhigh-resolution channel combined mechanism is designed, and the network performance is improved.

Description

technical field [0001] The invention belongs to the field of deep learning and medical image processing, and relates to data cleaning technology and convolutional neural network technology of thyroid nodule ultrasound image data sets, and in particular to a thyroid nodule classification method based on multi-scale feature fusion. Background technique [0002] Thyroid nodules are common diseases of the endocrine system, and 18% of adults carry thyroid nodules. Although the vast majority of thyroid nodules are benign, 10% of patients have malignant thyroid nodules, mainly thyroid cancer. In recent years, the incidence of thyroid cancer has risen rapidly, and currently ranks seventh in the incidence of cancer in the country. Ultrasound diagnosis is a common method to check benign and malignant thyroid nodules. However, because doctors generally rely on subjective judgments and lack objective standards, errors are prone to occur. [0003] The breakthrough of deep learning, es...

AI Technical Summary

Problems solved by technology

The above methods ignore the importance of end-to-end training, and do not design the network structure for the characteristics of ultrasound images of thyroid nodules

In addition, due to the issue of data confidentiality, most of the researchers' ultrasound images are non-public, and there is an urgent need for public large-scale ultrasound image data sets of thyroid nodules to be used by researchers.

[0005] Through the search of published patent documents, no published patent documents similar to this patent application have been found

Images