CAD of Brain Abnormalities in MRI Images Using Texture Features

The human brain plays a crucial and intricate role within the body, overseeing and harmonizing the functions of every organ. Any issues affecting the brain can result in disruptions to the body’s normal operations. A particularly perilous and prevalent concern today is brain tumors. Timely detection of brain tumors is imperative as it can improve a patient’s chances of survival. However, due to a rapid surge in cases over recent decades, manually identifying brain tumors from MRI images has become arduous and relies heavily on the judgment of Radiologists or Neurologists. To address the challenges of manual segmentation and minimize human intervention, we propose an automated approach for extracting diverse texture features from both normal brain and Glioma in MRI images. Comparing filter methods, the median filter demonstrates superior performance over the Gaussian filter, while the watershed algorithm outperforms the K-means clustering algorithm with higher Jaccard index and dice coefficient values. Features with a p-value below 0.05 are deemed significant, and these noteworthy attributes are subsequently employed to train a classifier aimed at detecting Glioma in MRI images.