Method for generating thyroid nodule classification model

Abstract

The invention provides a method for generating a thyroid nodule classification model, and the method comprises the steps: S1, extracting a plurality of thyroid nodule images from a sample database for each nodule type, and adding the images into a corresponding sample subset; S2, constructing a thyroid nodule classification network according to the nodule types, wherein the thyroid nodule classification network comprises a plurality of feature networks, and the feature networks are in one-to-one correspondence with the nodule types; S3, training the feature networks with the same nodule type according to the sample subsets to obtain network parameters corresponding to the feature networks, and configuring the network parameters in a thyroid nodule classification network; and S4, training the thyroid nodule classification network according to the thyroid nodule images in all the sample subsets, keeping network parameters unchanged in the training process, and obtaining a thyroid nodule classification model after the training is completed. The method has the advantage that the problems that an existing thyroid nodule model is too low in recognition accuracy and too high in false positive rate are solved.

Description

technical field [0001] The invention relates to the technical field of medical image processing, in particular to a method and system for generating a thyroid nodule classification model. Background technique [0002] Thyroid nodules refer to lumps in the thyroid gland, which can move up and down with the thyroid gland when swallowing. It is a common clinical disease and can be caused by various etiologies. Most thyroid nodules are benign, but a small number of thyroid nodules are malignant from the beginning, and some nodules will change from benign to malignant. The proportion of thyroid nodules turning into cancer is 5%, and if it can be detected and treated as soon as possible, most patients can survive for a long time. Currently, the examination of thyroid nodules mainly relies on ultrasound. The advantage of ultrasound is that it is cheap, flexible, real-time, and has no side effects. However, its disadvantages are that the image is unclear and the contrast is relativ...

AI Technical Summary

Problems solved by technology

[0003] The existing artificial intelligence recognition model of thyroid nodule ultrasound puts all types of nodule images together during training, and at the same time serves as the training sample set of the artificial intelligence recognition model, and uses the same feature extraction network and feature decoding network And the loss function, such a method, the recognition accuracy of the artificial intelligence recognition model obtained is low, and the false positive is high

However, due to the shape, size, and texture distribution of thyroid nodule images, when all types of thyroid nodule sample images are directly put into the model for training, it is not easy for the model to learn the eigenvalues ​​of each type of nodule at the same time.

If a smaller network model is used, the accuracy of model recognition will be too low; if a larger network model is used, the false positive rate of the model will be too high, and the thyroid nodule recognition model obtained by using the existing model network structure cannot Meet the needs of clinical diagnosis of thyroid nodules

Images