Thyroid cancer ultrasonic image automatic labeling method and system

Abstract

The invention discloses a thyroid cancer ultrasonic image automatic labeling method and system. The method includes the following steps that: a cancer image data set to be processed is preprocessed, so that the ROI sub-image data set of each cancer image is extracted; a VGG16 deep learning network model is adopted to perform feature extraction on the ROI sub-image data set; a K-means++ algorithm is used to cluster extracted features; clustering results are compared with the reference clustering results of preset cancer-free images, so that the cancer feature clusters of the cancer images are obtained; and the obtained cancer feature clusters are correspondingly labeled on the original images of the cancer images. The method and system of the invention have the advantages of high work efficiency, high accuracy and low application cost. The method and system can assist in saving a large quantity of financial and material resources and can be widely used in the medical image data processing field.

Description

technical field [0001] The invention relates to the field of medical image data processing, in particular to an automatic labeling method and system for ultrasound pictures of thyroid cancer. Background technique [0002] With the rapid development and maturity of computer storage capacity and computing power, related technologies of artificial intelligence have been greatly developed, especially related technologies of computer vision and natural language processing. At the same time, artificial intelligence is also constantly participating in different fields, improving the production efficiency and work efficiency of related industries in various fields, including the combination of medicine and artificial intelligence. [0003] At present, the combination of artificial intelligence and medicine is mainly reflected in the assistance of machines to doctors in diagnosis. Through artificial intelligence, using computer vision technology and deep learning, machines can assis...

AI Technical Summary

Problems solved by technology

However, due to technical limitations, at present, when training artificial intelligence to complete the recognition of thyroid cancer ultrasound pictures, it is mainly to select a large number of cancer pictures as the training set for training. This method has the following problems: 1) Training the model requires a large number of Ultrasound pictures of thyroid cancer are used as training samples; 2) The training set used to train the model requires a doctor to manually mark the cancer area on each picture

This method of marking a large number of pictures by doctors will undoubtedly consume a lot of time, and at the same time, a large backlog of doctors' time will be accumulated, resulting in a waste of hospital resources.

In addition, in this way, when it is necessary to continuously optimize the judgment accuracy of the model, the number of pictures that need to be marked will continue to increase. This method is inefficient and requires a lot of manpower and material resources, which is of little practical significance

Images