Automatic detection of disease from medical data is one of the common tasks in hospitals and recent approaches utilize the support from the deep learning (DL) schemes to enhance the disease detection accuracy. The proposed research aims to develop a DL-based scheme to detect the brain tumor (BT) from the chosen MRI database. The developed scheme consists the following phases; labeled data collection and resizing, feature extraction using chosen DL-model, feature optimization using Elephant-Herd Algorithm (EA), and classification and threefold cross validation. In this work, the VGG16-based scheme is executed to detect the BT using the conventional and optimized features. The experimental study is implemented using the Python-software and the outcome of this study confirms that the optimal feature-based BT-detection provides an accuracy of 98% with the K-Nearest Neighbor Classifier. This confirms the clinical importance of the developed technique.