Cascaded U-N Net brain tumor segmentation method combined with wavelet transform

Abstract

The invention discloses a cascaded UN Net brain tumor segmentation method combined with wavelet transform. The method comprises the following steps: S1, preprocessing an input image sequence; wherein the image sequence comprises four modal data FLAIR, T2, T1 and T1C and annotation data; s2, establishing a first layer of a cascaded network by using a Pytch deep learning framework, training a UNet model combined with wavelet transform, and then multiplying a training result of the model by the modal data T1 and T1C to generate a coarse segmentation image; and S3, establishing a second layer of the cascaded network by using a Pytch deep learning framework, training the NNet model combined with wavelet transform, and generating a fine segmentation image, namely a final brain tumor segmentation result. According to the method, the segmentation efficiency can be improved, and the problems of fuzzy tumor boundary, low segmentation precision and the like in brain tumor MRI segmentation can be solved.

Description

technical field [0001] The invention relates to the technical field of medical image segmentation, in particular to a cascade U-NNet brain tumor segmentation method combined with wavelet transform. Background technique [0002] Brain tumor segmentation is one of the branches of medical image processing. It is characterized in that according to the specific information such as texture and gray level difference between the brain tumor and other healthy brain tissues in the image, the brain tumor is separated from the image by using an algorithm. Separate or mark out. [0003] At present, the common type of brain tumor is glioma, and the composition of this type of tumor is generally composed of tumor nucleus, necrosis and edema. Differently, FLAIR and T2 modalities can highlight the edema area, while T1 and T1CE are sensitive to the information of tumor nucleus and necrosis. Accurate segmentation of brain tumor MRI is of great significance for improving tumor diagnosis, surg...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art, to provide a method for segmenting brain tumors with cascaded U-N Net combined with wavelet transform, aiming to use the method of cascaded networks to achieve step-by-step segmentation, and to solve the blurring of tumor boundaries in MRI segmentation of brain tumors , low segmentation accuracy and other issues; at the same time, the lightweight design of wavelet transform is introduced to reduce the parameter amount of the cascaded network and improve the segmentation efficiency

Images