Efficient Brain Tumour Classification Using Pre-Trained Xception Model

A brain tumour is among the most lethal tumours worldwide. It exhibits the lowest survival rate and presents various varieties based on location, texture, and form. Thus, accurately determining the type and grade in the early stages is crucial to selecting an appropriate treatment strategy. The substantial volume of data and the diverse forms of brain tumors render the manual procedure time consuming and prone to human error. This research study presents a brain tumour classification using the Xception transfer learning model. The proposed Xception model is trained and tested using the Brain Tumour Image Dataset collected from Kaggle. The raw data are preprocessed before applying to transfer learning model to match its parameters. The proposed model provides a testing accuracy of 0.98 and a loss of 0.02 and shows exceptional efficiency contrasted to the existing models. In real-time, this model assists the doctor in identifying the particular brain tumor.