Deep Transfer Learning Models for Brain Tumor Classification Using Magnetic Resonance Images

The development of brain tumor cells causes the intracranial pressure inside the skull to rise, which ultimately threatens the minuscular life cycle. So, medication is the only way to increase life expectancy. It may cause death if not treated in the beginning stages. In this research paper, we aim to increase the effectiveness of magnetic resonance imaging (MRI) devices to identify cancerous brain tumor cells. We classify brain tumor cancer cells using five pre-trained convolutional neural network architecture models: ResNetl52, DenseNet201, VGG19, MobileNetV2, and InceptionV3. The ResNetl52 model was found to have the highest accuracy rate after the work was analyzed, with an accuracy value of 98.52%, which is almost 99%. In addition, the VGG19 model also showed a good accuracy rate of 98%. Other models have accuracy values of InceptionV381.30%, DenseNet20183.30%, and MobileNetV286.73%, respectively. Lower ranking means models are unable to identify brain tumor cancer cells compared to other models in the same magnetic resonance image. Based on the result, it has been observed that ResNetl52 is suitable for brain tumor cancer cell detection. The early diagnosis of malignancies before they develop physical adverse effects is benefited from this accuracy.