Image Processing Based Detection of Abnormal Tissue in Magnetic Resonance Images

Biomedical imaging methods facilitate easy visualisation as well as recognition of the slightest abnormalities within the human body. Cognizance of abnormalities before they turn detrimental plays a deterministic role in preserving the quality of human health along with enhancing life expectancy and longevity of life. Early diagnosis of brain tumour is an arduous task as discernible physical symptoms appear in the advanced stages of tumour. On the other hand, MRI provides an expeditious and accurate insight into the condition of tumours in advance to support its treatment at a preliminary stage itself and is therefore most widely preferred in the medical community. Implementation of image processing techniques on biomedical DICOM images assist in the detection of even the most minute cell abnormalities with faster speed, higher efficiency and low probability of human error. In this paper, a novel framework has been designed for the identification of abnormal tissue in MRI images using texture feature extraction methodology. Five texture features namely contrast, correlation, energy, homogeneity and entropy have been extracted from the GLCM matrix of the sample of tumour images and used as input to a Support Vector Machine. The low values of misclassification error in SVM and p-values in ANOV A are clearly indicative of the efficiency and accuracy of the algorithm.