Open AccessBrain Tumor Classification Using Texture Feature Extraction
Author(s) -
Heba Kh. Abbas,
Nada A. Fatah,
Haidar J. Mohamad,
Ali A. Al-Zuky
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1892/1/012012
Subject(s) - artificial intelligence , pattern recognition (psychology) , grey level , histogram , computer science , segmentation , feature extraction , image texture , homogeneity (statistics) , computer vision , feature (linguistics) , image segmentation , image processing , energy (signal processing) , image (mathematics) , mathematics , statistics , linguistics , philosophy , machine learning
Medical imaging methods are necessary for tumor treatment. Image processing increases the accuracy, location, efficiency and recognizes tumor in magnetic resonance imaging (MRI) for the human brain. There are three MRI image cases studied consist of one image of a healthy human brain, and two with brain cancer cases to extract tumor size, area, dimension, and location. We presented two steps to extract the texture properties from the three image cases. The first step consists of two methods, used to extract statistical image feature, which is a grey level co-occurrence matrix modified method and the proposed grey level Run-length matrix method. The second step consists of two methods applied to the outputs of step-one to extract the full information. These methods are segmentation based on threshold technique and supervised classification technique based on the parallel pipe. Moreover, there are twelve features studied like the image roughness, uniformity or energy, and local homogeneity extracted to show the quality difference between methods. Image histogram presented to show the small difference between the output of the features. The result shows the segmentation based on threshold technique has efficient output with higher accuracy as well as the presented methods.