Thyroid slice image classification model training method and device

Abstract

The invention relates to a thyroid slice image classification model training method and device. The method comprises the following steps: acquiring a thyroid slice image under a preset magnification; dividing the thyroid slice image into a plurality of image blocks which are not overlapped with each other and have a preset size; classifying the image blocks through an image block classification model to obtain the probability that the image blocks are malignant tumors; in the training process of the image block classification model, using an optimization loss function, and adjusting model parameters through a back propagation algorithm; mapping the probability to the position of the image block corresponding to the probability in the thyroid slice image to obtain a probability heat map of the thyroid slice image; and extracting a characteristic value of a tumor from the probability heat map, inputting the characteristic value into an SVM classifier for training, and obtaining a thyroid slice image classification model. By adopting the method, the model classification accuracy can be improved.

Description

Technical field [0001] The present application relates to the field of image processing, and in particular, to a thyroid sewer image classification model training method, a device. Background technique [0002] The global incidence of thyroid cancer has continued to rise in recent decades, and the growth rate is ranked in all solid tumors, and the incidence is increasing at 6% per year. At present, thyroid cancer is still subject to surgical treatment. Since the frozen slices in thyroid are highly specific, it is often conventionally conventional frozen slices to determine further treatment regimens. However, there is a great challenge to find a tumor region and judge the sliced ​​of the sliced ​​of the sliced, and the pathologists may interpret the thyroid frozen slices because of insufficient or fatigue. Missed the best treatment period. Therefore, the depth learning application can be used in terms of thyroid freezing slices, and the auxiliary doctors can be diagnosed, which c...

AI Technical Summary

Problems solved by technology

However, most of the current advanced classification networks have serious misdiagnosis when classifying image blocks. This is because of the diversity of cell morphology in thyroid slices, and some benign tumor areas and even normal areas and malignant tumor areas have similar visual features.

It is precisely because of the diversity of cell morphology that the image block classifier has a serious underdiagnosis situation, which indirectly leads to a serious underdiagnosis situation in the whole slice classification.

However, it is not allowed to judge malignant thyroid tumors as benign thyroid tumors or even normal tissues in clinical diagnosis, because such missed diagnosis will directly cause patients to miss the best treatment period

[0004] In short, in the prior art, the method of judging benign and malignant thyroid tumors based on thyroid slice images is easy to judge malignant tumors as benign tumors with a higher probability

Images