Unveiling the Hidden: Leveraging Medical Imaging Data for Enhanced Brain Tumor Detection Using CNN Architectures

Abstract

Brain tumor detection using deep learning has made significant progress, but there are still several challenges and problems that researchers and practitioners are actively addressing like limited data availability, imbalanced data, generalization of data, data preprocessing and integration into clinical practice.To overcome these challenges, this study proposes the use of several transfer learning techniques and CNNs to provide a unique method for classifying brain tumors.Specifically, we employed three well-known transfer learning architectures, namely VGG, ResNet, and MobileNet, to explore their performance in brain tumor detection.Advantages of using VGG, ResNet, and MobileNet models include their ability to leverage pre-trained knowledge, adaptability to different problem domains, architectural diversity, simplicity, efficiency, state-of-the-art performance.Deep learning and models with prior training are used to improve the accuracy and efficiency of classifying brain tumors.The comparative study of various models showed that in order to classify brain tumor images, MobileNet stands out with the highest accuracy of 98.66% as compared to 97.55% of VGG and 87.44% of ResNet.The outcomes of this project help advance the field of diagnostic imaging and aid medical practitioners in the prompt and precise identification of brain tumors.