MRI brain tumor localization and intratumoral segmentation method based on deep cascaded convolutional network

Abstract

The invention provides a MRI brain tumor localization and intratumoral segmentation method based on a deep cascaded convolutional network, comprising the following steps: building a deep cascaded convolutional neural network segmentation model; model training and parameter optimization; multimodal MRI brain tumor Rapid localization and intratumoral segmentation. The MRI brain tumor location and intratumoral segmentation method based on the deep cascaded convolutional network provided by the present invention constructs a deep cascaded hybrid neural network composed of a full convolutional neural network and a classification convolutional neural network, and divides the segmentation process into complete Tumor area positioning and intra-tumor sub-area segmentation are two stages to achieve rapid and accurate hierarchical MRI brain tumor positioning and intra-tumor sub-area segmentation. First, the full convolutional network method is used to locate the complete tumor area from the MRI image, and then the image block classification method is used. The complete tumor is further segmented into edema area, non-enhancing tumor area, enhancing tumor area and necrosis area, realizing the precise localization of multimodal MRI brain tumors and the rapid and accurate segmentation of intratumoral sub-regions.

Description

technical field [0001] The invention relates to the technical field of medical image analysis, in particular to an MRI brain tumor localization and intratumoral segmentation method based on a deep cascaded convolutional network. Background technique [0002] Brain tumor is a major disease that seriously endangers human health. Among them, glioma is the main type of malignant brain tumor. Although it is uncommon, it has a very high fatality rate. According to literature statistics, the average survival time of high-grade glioma is 14 months. Magnetic Resonance Imaging (MRI) is the most commonly used method for the examination and diagnosis of brain tumors in clinical practice. Accurately segmenting brain tumors and intratumoral structures from MRI images is of great value for neuropathological analysis and accurate diagnosis. Provide important support for surgical planning, radiotherapy and chemotherapy plan formulation and prognosis assessment. [0003] The segmentation o...

AI Technical Summary

Problems solved by technology

[0006] In the existing MRI brain tumor segmentation, the image block classification method does not use global context features and the segmentation speed is slow, while the training samples of the fully convolutional network pixel-level classification method are seriously unbalanced, resulting in inaccurate segmentation boundaries of small regions. The present invention provides a An MRI brain tumor localization and intratumoral segmentation method based on a deep cascaded convolutional network. This method divides the segmentation process into two stages: complete tumor area localization and intratumoral sub-area segmentation, by building a deep cascaded hybrid convolutional neural network. Fast and accurate localization and intratumoral subregion segmentation of multimodal MRI brain tumors

Images