Due to the large number of MRI slices provided for each patient, the MRI visual evaluation process of brain disease diagnosis has become time-consuming, slow and more vulnerable to error. In this research an algorithm is proposed to extract texture feature and classify MRI brain scans into normal and abnormal MRI scans. First, the MRI scans are pre- processed by image enhancement, intensity normalization and correcting the mid-sagittal plane (MSP) of brain. Second, the proposed bi-directional modified gray level co-occurrence matrix (Bi-MGLCM) method is used to extract texture features from MRI T2-weighted images that are used to measure the degree of symmetry between the left and right hemispheres of the brain. Finally, these features are classified into normal and abnormal by using longshort term memory (LSTM) model. The research will be validated using two datasets; a real dataset that was collected from the magnetic resonance imaging unit in Al-Imamain Al-Kadhimain Medical City – Baghdad - Iraq in 2019 and the standard data set for classification of brain tumors (BRATS 2013). The achieved classification accuracies were 96.3% for the collected dataset and 98.9 % for the BRATS 2013.

Deep Learning, MGLCM, MRI Brain Scans, Feature Extraction, LSTM Classifier.