Optimized Brain Tumor Detection and Classification on Efficient Net B3

Magnetic Resonance Imaging (MRI) is a medical imaging technique used for multiple reasons especially when it comes to Brain Tumor. MRI has helped physicians to detect and classify the types of tumors but then it is quite difficult to classify the image due to lack of details and expertise from physicians. In this paper, an attempt has been made to classify the type of tumor if at all tumor is present and if not then it detects that tumor is not present with the help of a neural network which is lightweight and has fewer floating point operations. The MRI scan was processed by histogram equalization and then contour detection which allows the important details to be amplified and which in turn allows the CNN model to work in a better fashion. For this Efficient Net B3 has been used as a training model which is chosen due to its robustness and yet lightweight. The hyperparameter values for the model have been optimized using Optuna a Python-dependent software. The model is then evaluated and performance is tested on several data.