Deep Learning based Autonomous Model for Detection and Classification of Brain Tumor Disease

Due to the inflammation, injury, poor diet, stress and any other genetic issue, abnormal tissues can grow inside brain and cause brain tumor if not diagnosed and treated timely. Though many medical experts prefer magnetic resonance imaging (MRI) and computed tomography (CT) scans for identifying presence of tumor which signifies that interpretation of such traditional approaches completely depend up on expertise of radiologists who can also give inaccurate predictions especially at initial stage. Therefore, developing autonomous system in medical field to diagnose presence of tumor has become necessity to aid experts in making accurate and quick decision. To give significant contribution in this regard, data has been collected from publicly available source. Afterwards, residual network with 101 layers has been used for extraction of significant features following transfer learning and linear support vector classifier is used for distinguishing infected images from normal ones. Therefore, such integrated model prevents from overfitting issues and minimizes time-complexity due to which this model outperformed the existing techniques. Moreover, the proposed model is validated on unseen data to know its effectiveness. Based on testing data, hyperparameters and other performance-metrics from confusion-metrics have been obtained. Thus, model gained 96.33% accuracy, 97% sensitivity, 95.67% specificity, 95.72% precision and 96.36% F1-score. Thus, achieved performance-metrics showed efficient and improved outcomes as compared with state-of-art approaches.