Advancing Brain Tumor Diagnosis with Convolutional Neural Networks: A Multi-Class Classification Study

Advancing Brain Tumor Diagnosis with Convolutional Neural Networks: A Multi-Class Classification Study

Deepika Sharma, Yash Puri, Shivam Gariya, Raj Vishwakarma, Dhruv Narayan, Swastik Sharma

Department of Computer Science, Chandigarh University, Mohali, Punjab, India

Abstract— Tumours that occur in the brain are dangerous and can either be benign or malignant; therefore early detection usually determine the effectiveness of treatment. Our contribution in the present paper is the proposal and evaluation of a deep learning model based on CNNs for the classification of brain tumors using MRI. Based on a database of 7,023 MRI images with glioma, meningioma, pituitary and no (normal) tumor, we formulated a multi-class classifier that recognizes the difference between these tumors. The input data was preprocessed to correct their size and normalize them before feeding them to the model; the data was augmented as well to improve model performance and prevent it from memorizing the data points. Therefore, the training was done using the CNN model together with the Adam optimizer, tuning down the hyperparameters to suit the highest performance. All model assessment used accuracy, precision, recall, and F1 score for efficiency on the various classes. A test accuracy of our model was 99.14 %, with precision and recall values higher than 98 % for all the tumor classes.

The outcomes also clearly establish the viability of CNN- based solutions in aiding the process of initial stage diagnostics and therapy of brain tumours. Future work could involve using the model to analyze even bigger population and including it in clinic for immediate diagnosis.

Keywords— Brain Tumor Classification, Convolutional Neural Networks, MRI Images, Deep Learning, Glioma, Meningioma, Pituitary Tumor, Medical Imaging, Multi-Class Classification, Tumor Diagnosis.