Brain glioma segmentation based on cascaded convolutional neural network

Abstract

The invention discloses a brain glioma segmentation method based on a cascaded convolutional neural network, and the method comprises the steps: carrying out the primary coarse segmentation of a braintumor region, and extracting the approximate position information of a tumor; expanding 10 pixels for each dimension on the basis of coarse segmentation and taking the 10 pixels as input of a fine segmentation network; improviing the fine segmentation network, so as to enable the fine segmentation network to combine the advantages of dense connection, an improved loss function and multi-dimensional model integration; designing an integrated model of three directions (2D, 2.5 D and 3DCNN models), and respectively considering all information of different resolutions corresponding to each direction; integrating post-processing operation condition random fields in a segmentation algorithm, and optimizing continuity of segmentation results in appearance and spatial positions. According to themethod, the brain glioma is segmented through the two-step cascaded convolutional neural network, the advantages of dense connection, a new loss function and multi-dimensional model integration are combined, an integration model in multiple directions is designed, and finally a segmentation result is optimized through a conditional random field.

Description

technical field [0001] The invention relates to brain glioma segmentation based on a cascaded convolutional neural network, and belongs to the field of medical image processing. Background technique [0002] Glioma is the most common primary malignant brain tumor. Clinically, doctors mainly analyze tumor images to formulate treatment plans for patients and evaluate treatment effects. Some non-invasive and easy-to-obtain biomarkers can be obtained from medical images to describe tumor status and treatment response, such as tumor contour features, border texture, cross-sectional area and volume, etc., which are used by doctors in formulating treatment plans. necessary reference factors. The first step of tumor classification is to accurately segment tumors of different shapes, but the current segmentation work is mainly done manually, which is time-consuming and laborious, and may lead to the loss of useful information. Therefore, automatic and accurate segmentation of brain ...

AI Technical Summary

Problems solved by technology

These two types of methods have their own advantages and disadvantages. Although 3D-CNN can make full use of the potential 3D information of MRI data, it will also increase the network size and computing cost (higher hardware requirements) and it is difficult to find pre-trained Model (the amount of data in 3D datasets is relatively small) and other issues

Images