Investigation and Classification of MRI Brain Tumors Using Feature Extraction Technique

Medical imaging is a novel research area in the domain of image processing for the research community. Features computed from MRI images provide a high level of information used in medical diagnostics. This paper addresses the classification of different types of brain tumors studied in MRI images using feature extraction techniques. It may help in the effectiveness of brain tumor treatment that depends on the early detection needed to distinguish between benign and malignant tumors. We present in this paper, a novel framework to investigate and classify brain tumors in DICOM format T2-FLAIR MRI images. Spatial filters are used to remove undesired information and noises. Segmentation is done using a thresholding method to separate the tumorous regions from healthy regions. Then, the Discrete Wavelet Transform is employed for reducing the dimensionality of the images followed by Principal Component Analysis, which reduces the dimensions further while keeping the only useful information. The gray level co-occurrence matrix is implemented to extract the texture features that can be fed into a Support Vector machine for further classification. The proposed framework is able to distinguish benign and malignant tumors precisely. The extracted features show that the benign tumor is more homogeneous with higher energy than malignant and have lower values of contrast, correlation, and entropy than malignant. The results also illustrate that benign tumors have more irregular appearance than malignant tumors. The accuracy of the proposed method is 95%. The errors in the segmentation and high dimensionality of the DICOM images causes some lack of accuracy. The extracted features demonstrate that the proposed method can be used to investigate and classify the types of brain tumors. In future work, it can also be tested for tumor detection of other parts of the body. The proposed framework can be used as a quick guidance tool for the radiologists.