A transfer learning-based brain tumor classification using magnetic resonance images

The brain is one of the most important and complex organs responsible for controlling the functions of the human body. Brain tumors are among the most lethal malignancies in the world. Analyzing MRI images of the patient's brain is one of the usual approaches for diagnosing brain tumors. However, misdiagnosis of brain tumor forms hinders effective responses to medical interventions and reduces patients' chances of survival. With recent technological advancements, deep learning-based medical image analysis has become increasingly popular in the research community. Transfer learning reuses a pre-trained model, trained on huge volume datasets to solve a new problem in distinguished application domains. This research article attempts to diagnose the three most prevalent forms of brain tumors using pre-trained CNN models such as VGG19, Inception-v3, and ResNet50 using transfer learning. The features extracted using pre-trained models are supplied into fully connected layers that fine-tune the model for classifying multi-class tumors. The performance of the presented approach is evaluated on a benchmark MRI esults reveal that te proposed approach leads to superior performance in comparison to conventional techniques with an average accuracy of 90%.