A 3D multi-feature fusion tooth CT image segmentation method

Abstract

The invention discloses a tooth CT image segmentation method based on 3D multi-feature fusion, which belongs to the technical field of image processing. The present invention first carries out image conversion processing to CT image, converts it into grayscale image; Then builds the neural network model for tooth segmentation and trains it, and the skeleton of described neural network model adopts U-net network; Finally, The CT image preprocessing is performed on the image to be segmented, and the gray image is obtained and input to the trained neural network model, and the segmentation result is obtained based on its output. The present invention combines the similarity between the top and bottom of CT images, and proposes a 3D multi-feature fusion tooth segmentation method, which improves the training accuracy of the neural network; at the same time, it combines the CRF algorithm to remove redundant information generated by the neural network model, making the segmentation results more accurate. accurate.

Description

technical field [0001] The invention relates to the technical field of medical CT (Computed Tomography, computerized tomography) image segmentation, in particular to a tooth CT image segmentation method based on neural network-based 3D multi-feature fusion. Background technique [0002] Medical CT image segmentation has great value in clinical applications, but compared with natural images, medical CT images have many defects: (1) low resolution and various artifacts on the image increase the difficulty of segmentation, (2) corresponding Tissues are difficult to calibrate. Today's segmentation methods are mainly divided into: neural network methods and non-neural network methods, and neural network methods are further divided into 2D neural network methods and 3D neural network methods. [0003] Non-neural network methods mainly use random forests for cascade segmentation (HAAR feature segmentation, HOG feature segmentation) and use Level Set (level set) for edge segmentati...

AI Technical Summary

Problems solved by technology

[0002] Medical CT image segmentation has great value in clinical applications, but compared with natural images, medical CT images have many defects: (1) low resolution and various artifacts on the image increase the difficulty of segmentation, (2) corresponding Tissue is difficult to calibrate

[0006] Two-dimensional neural networks have made great progress in medical image segmentation, but it is difficult to further improve, because 2D neural networks have some limitations: (1) The training set labels used by 2D neural networks are difficult to distinguish information between similarly located tissues

(2) The 2D image itself is difficult to accurately express the differences of different individuals

Good results can also be achieved in medical image segmentation, but due to the characteristics of medical images, this method still has many deficiencies and needs further improvement

Images