An Interpretable Brain Tumour Classification Approach using Machine Learning and CNN-based features

Cancer, a disease that does not discriminate; impacting people from all walks of life. Numerous types of cancer affect humans, each with distinct characteristics and treatment approaches. One of the most effective strategies in combating cancer is early detection. Recently, several learning algorithms have been implemented to classify and identify brain tumors. However, a significant challenge with these models is their lack of interpretability. Explainable AI models like SHAP and LIME are apt and capable of clearly providing accurate reasoning to the models. Our research dwells on the methodology of brain tumor classification. This paper focuses on leveraging learning models by exploring the use of machine learning models for their efficient classification capabilities and deep learning architectures for their ability to detect subtle features. The study is based on a dataset of 7,000 MRI scans of brain tumors distributed into 4 different classes. We learned that the SVM classifier had a better performance of 97%. We were able to shed light on the accuracy and working with the help of SHAP plots and LIME heat maps.