Cascade neural network structure-based brain glioma segmentation method

Abstract

The invention provides a brain glioma segmentation method based on a cascade neural network structure. The method comprises the following steps: 1, generating a high-precision tumor region segmentation result by using a brain glioma segmentation network of the cascade neural network structure; 2, aiming at the multi-scale residual features and the global features, on one hand, generating a whole tumor segmentation result and an edge detection result by utilizing a segmentation and edge detection network; on the other hand, a cascade network is designed to generate tumor core region and tumor enhancement region segmentation results under the preliminary whole tumor segmentation result; 3, constructing a loss function to train the precise brain nerve tumor segmentation network; and outputting: performing tumor region segmentation on the original multi-modal image by using the trained brain glioma segmentation network of the cascade neural network structure. The method can be combined with various medical image-based application systems, helps to improve the segmentation quality of multi-modal images, and has wide market prospects and application values.

Description

technical field [0001] The invention relates to a brain glioma segmentation method based on a cascaded neural network structure, which belongs to the fields of digital image processing, pattern recognition and computer vision. Medical image segmentation has broad application prospects in various image-guided interventional diagnosis and treatment and targeted radiotherapy application systems. Background technique [0002] Gliomas are the most common primary brain malignancies with varying degrees of aggressiveness, prognosis, and regional heterogeneity. Segmentation of glioma usually refers to segmentation of tumor regions from multimodal MRI sequences. The segmentation of glioma can effectively extract multiple heterogeneous regions of the tumor (including the entire tumor region, tumor core region and tumor enhancement region), thereby helping doctors to make accurate judgments. Medical image segmentation is more challenging than ordinary color images due to various nois...

AI Technical Summary

Problems solved by technology

[0006] However, the current medical image segmentation methods based on convolutional neural networks are mostly based on the grayscale information of the image to extract the region of interest, which does not make full use of edge information and lacks the consideration of the spatial relationship between classes in multi-class segmentation tasks, resulting in insufficient 3D segmentation. , the accuracy is not high

Images