A method for automatic segmentation of brain tumor images based on superpixel fuzzy spectral clustering

Abstract

The invention discloses a method for automatic segmentation of brain tumor images based on superpixel fuzzy spectrum clustering. Firstly, superpixel segmentation is performed on the FLAIR modal images of nuclear magnetic resonance imaging containing brain tumors, and then the gray level histograms of these superpixel blocks are extracted. Graph feature, the gray histogram feature of the superpixel block is used as the input of the algorithm, the fuzzy similarity matrix of the image is calculated through the input feature, and then clustered by the NJW spectral clustering algorithm to obtain the final segmentation result. The invention optimizes the similarity measure method of spectral clustering by using fuzzy theory, introduces fuzzy weight parameters into the Gaussian distance measure method of spectral clustering, and defines a fuzzy similarity measure method based on superpixel features. The invention is an automatic image segmentation method without human intervention, and utilizes a fuzzy spectral clustering segmentation algorithm based on superpixels to greatly reduce the time complexity of the spectral clustering algorithm and improve segmentation accuracy.

Description

technical field [0001] The invention relates to the fields of image processing, pattern recognition and machine learning, in particular to a brain tumor image segmentation method based on fuzzy clustering. Background technique [0002] Brain tumor is a disease with a high incidence rate in modern times. This disease can oppress the normal brain tissue of the human body, which has a great impact on the health of patients, and may even lead to death. At present, the incidence of brain tumors accounts for 1.5% of tumors in the whole body. In recent years, with the wide application of imaging technologies such as magnetic resonance imaging and CT, it provides a basis for doctors' reliable diagnosis. Because such techniques are relatively simple, fast and effective, and have high accuracy, the detection rate of brain tumors has been greatly improved. At present, the treatment plan for brain tumor diseases is basically based on surgery, supplemented by radiotherapy and chemothera...

AI Technical Summary

Problems solved by technology

Among them, manual segmentation requires doctors to have rich clinical experience, and the disadvantages are obvious. For example, the segmentation process is very time-consuming, and the segmentation results of different experts are different, even if the same expert has different results in different time periods; semi-automatic segmentation is often The complete segmentation process can be realized by means of "human-computer interaction" such as initialization operation or manual correction of segmentation results. At present, this semi-automatic segmentation method is widely used in clinical applications; automatic segmentation usually combines human intelligence and prior knowledge, using software Computing and other technologies realize intelligent segmentation without human intervention. However, due to the diversity of brain tumors in different patients and the characteristics of easy confusion with normal tissues in most cases, the accuracy of automatic segmentation has always been a challenge.

Images