An Ensemble of Optimal Deep Learning Features for Brain Tumor Classification

Owing to technological developments, Medical image analysis has received considerable attention in the rapid detection and classification of diseases. The brain is an essential organ humans. Brain tumors cause loss memory, vision, name. In 2020, approximately 18,020 deaths occurred due tumors. These cases can be minimized if a tumor diagnosed at very early stage. Computer vision researchers have introduced several techniques for classification. However, owing many factors, this still challenging task. challenges relate size, shape tumor, location selection important features, among others. study, we proposed framework multimodal using ensemble optimal deep learning features. framework, initially, database normalized form high-grade glioma (HGG) low-grade (LGG) patients then two pre-trained models (ResNet50 Densenet201) are chosen. were modified trained transfer learning. Subsequently, enhanced ant colony optimization algorithm best feature from both models. selected features fused serial-based approach classified cubic support vector machine. experimental process was conducted on BraTs2019 dataset achieved accuracies 87.8% 84.6% HGG LGG, respectively. comparison performed methods, it shows significance our technique.