Symmetric brain tumor segmentation method based on neural network

Abstract

The invention provides a symmetric brain tumor segmentation method based on the neural network. The method mainly comprises symmetric coding of the neural network, a dataset, preprocessing and model comparison. A difference between an original brain image and an overturned brain image is calculated, features are extracted from the difference, the difference is connected with a characteristic graphof the original image to reserve information of the original image, and symmetry in the present neural network is coded. The brain symmetry is highly abstract, the symmetry is subtracted after convolutional and pooling layers, and characteristic of different abstract layers are extracted from the neural network with continuous convolutional and pooling layers. Thus, the problem that a brain tumoris complex in structure, fuzzy in boundary and mixed and overlapped with normal brain tissues is solved, the asymmetry of the brain is focused in segmentation, problems in brain tumor segmentation isovercome, and the correct rate of brain tumor segmentation is improved.

Description

technical field [0001] The invention relates to the field of image segmentation, in particular to a neural network-based symmetric brain tumor segmentation method. Background technique [0002] Image segmentation is the technology and process of dividing an image into several specific regions with unique properties and proposing objects of interest. Brain tumor images, as a special medical image, also belong to natural images. Brain tumor image segmentation algorithm is the application of natural image segmentation algorithm in brain tumor images. At the same time, it needs to consider the particularity of brain tumor images. At present, brain tumors have become an important disease that threatens human health. Every year, many people around the world suffer from Brain tumors have lost their lives, so the contour extraction of tumors and surrounding organs has very important research significance for the diagnosis and treatment of tumors. In clinical practice, accurate segm...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problems of brain tumors with complex structures, blurred borders and aliasing with normal brain tissue, the purpose of the present invention is to provide a symmetrical brain tumor segmentation method based on neural networks. First, calculate the original brain image and the flipped brain The difference between images, extract features on the difference, connect the difference with the feature map of the original image to preserve the information of the original image, and then encode the symmetry in the existing neural network; since the brain symmetry is at a high level of abstraction, so Subtracting symmetries after several convolutional and pooling layers to extract features at different levels of abstraction on neural networks with successive convolutional and pooling layers

Images