Explainable Convolutional Neural Network for Brain Tumor Classification via MRI Images

Brain tumor classification is a crucial problem in the domain of Computer-Aided Diagnosis (CAD). In this thesis, we are focusing on the binary classification of brain images obtained from MRI scans of the human brain from different angles along with a proper explanation of the predicted results. The propose classification model is based on the concept of deep transfer learning using a pre-trained VGG16 model and “Image Net” weights to automatically extract features from the MRI brain image inputs. Here comes the role of explainability which will help us understand the decision-making process of our AI system to build trust and transparency. To reach this goal, we are using Shapley values to explain the whole model and provide a visualized explanation of the outcome.