Abstract
A tumor can be any extra and uncontrolled growth of cells in the body. The brain tumor is one of the deadliest diseases because of the structural complexity and functioning of the brain. It is crucial to identify the brain tumor at an initial stage. MRI is one of the popular imaging modalities used in the field of biomedical imaging as it provides good contrast images. The advancement in imaging using CAD and deep learning makes tumor diagnosis easy. This paper proposes a comparative study between two deep learning models: CNN and transfer learning-based deep learning models. A dataset from the publicly available platform Kaggle is used consisting of 5712 images. The input image is first pre-processed using a Gaussian filter, then segmentation is done by using snake segmentation. Then, the segmented image is again de-noised by using PNLM filters. In the end, both the classifier models are compiled, trained and compared in terms of accuracy, precision, recall and F1-score.