Oral cavity curved image wisdom tooth segmentation method based on YOLO and U-Net

Abstract

The invention discloses an oral cavity curved image wisdom tooth segmentation method based on YOLO and U-Net, and the method comprises the steps: 1, carrying out the image preprocessing of an oral cavity curved image, carrying out the position labeling of wisdom teeth in the preprocessed image, obtaining a position label, and dividing a training set and a test set; 2, training a YOLO model by using the image and the position label, setting a confidence coefficient threshold value of a spatial position, and obtaining spatial position information of the wisdom tooth; 3, carrying out slicing processing based on the spatial position information of the wisdom teeth, preprocessing all obtained slices containing the wisdom teeth, and carrying out pixel-level category labeling on the slices of the training set after preprocessing; and 4, training a U-Net model through the slice images and the position labels of the training set, setting a confidence coefficient threshold value of a pixel category, carrying out binaryzation, obtaining pixel-level classification information of the wisdom teeth, and completing wisdom tooth segmentation. The method achieves the wisdom tooth segmentation of the oral cavity curved image through two stages, effectively improves the positioning precision of the wisdom tooth, reduces the calculation cost of the wisdom tooth segmentation, and improves the speed and precision of the wisdom tooth segmentation of the oral cavity curved image.

Description

technical field [0001] The invention relates to the technical field of medical image segmentation, in particular to a method for segmenting wisdom teeth based on oral curvature images based on YOLO and U-Net. Background technique [0002] Teeth are one of the important organs of human beings, and the problems caused by wisdom teeth account for a considerable proportion of oral diseases. The growth status of wisdom teeth is also an important basis for surgical removal; judgment plays an important role. Therefore, compared with other teeth, choosing wisdom teeth as the research target has a wider audience and higher clinical application value. At the same time, the most common form of stomatology imaging in China is oral tortuosity image. [0003] In 2015, Long Jonathan et al. [1] proposed a fully convolutional network, which can classify images at the pixel level, thereby solving the problem of image segmentation at the semantic level. This work is the pioneering work of d...

AI Technical Summary

Problems solved by technology

[0005]However, U-Net is directly applied to the wisdom teeth segmentation of oral curvature images, and there are the following two problems: (1) In the oral curvature images with more complex features , because the structure of wisdom teeth is similar to other teeth but the spatial position is different, U-Net can capture the morphological characteristics of teeth, but it cannot accurately capture the spatial position characteristics of wisdom teeth, so it is impossible to obtain accurate wisdom teeth segmentation results simply by using U-Net; (2 ) The size of the oral cavity image is relatively large, while the proportion of wisdom teeth is relatively small. A single wisdom tooth only accounts for about 0.6% of the entire image. Using U-Net to directly segment the oral cavity image will result in a waste of most of the computing power. background area of ​​wisdom teeth

Images