A comparative study of brain tumor classification on unbalanced dataset using deep neural networks

A brain tumor can pose a significant risk to a person's life if it is not promptly detected and diagnosed. The transfer learning approach is used in the majority of research studies on the identification of brain tumors using magnetic resonance imaging (MRI). In this study, ten state-of-the-art deep learning models are evaluated and compared on an unbalanced dataset for three-class brain tumor classification. Experimentation results reveal that the Inception models outperformed all the models for the three-class classification. However, the EfficientNet model does not perform well in comparison to all other models. Moreover, this study provides insight into the well-known deep neural networks on the MRI dataset for brain tumor classification. This study will allow the readers to analyze the performance of these models provided through detailed tables, confusion metrics, and loss graphs in their future research.