Optimized Deep Neural Network for Accurate Detection of Malignant and Benign Brain Tumors

The rapid development of medical technology has ushered in a new age of crucial medical records. A brain tumor is an aberrant mass of rapidly expanding brain cells, some of which have the potential to metastasize. Brain tumors are often diagnosed in clinical practice using magnetic resonance imaging (MRI) scans. These tumors appear as clusters of irregular cells that could endanger patients by invading nearby tissues. However, manually spotting and segmenting brain tumors via MRI scans present complex challenges and are prone to mistakes. Thus, analysis of such data has influenced the prognosis, monitoring, diagnosis, and treatment of tumor-related conditions. To overcome these issues, we propose an Optimized Deep Neural Network (ODNN) method to accurately predict benign or malignant brain tumor diseases. Moreover, pre-processing uses an Adaptive Histogram Contrast Normalization (AHCN) technique to enhance the distribution and contrast of MRI image density. Afterward, features can be extracted using a Texture-Based Gray Level Co-Occurrence Matrix (TGLCM) method to separate the dark and bright parts of the image. Finally, a Machine Learning (ML) based ODNN method is employed to classify benign or malignant diseases. Utilizing sensitivity, specificity, recall, and accuracy metrics, the simulation results of the recommended ODNN approach are contrasted with those of preexisting approaches.