Elevating Brain Tumor Classification Accuracy with Pre-Trained CNNs and Transfer Learning from MRI Data

Brain Tumor are thoughtful condition initiated through uncontrolled and abnormal division of cell. If a tumor is not accurately and quickly diagnosed, the consequences can be disastrous. MRI is frequently used methods by doctors to detect brain tumor due to its excellent resolution and it can clearly find tiny, non-invasive abnormalities in the brain. One of the top causes of death in people of all ages is brain tumors. The importance of early detection and treatment in enhancing patient outcomes. For radiotherapy to be effective, it is essential to accurately determine the tumor's location, size, and dimensions. Despite its accuracy, manual tumor classification on MRI scans is time-consuming. Computerized methods can deliver more thorough and accurate results in less time to address this problem. This study aimed to explore the potential of pre-trained Deep Convolutional Neural Networks (DCNN), namely VGG-19, VGG-16, CNN in categorizing brain MR images by employing transfer learning techniques. Assessing the test set's validation through metrics like accuracy, recall, precision, and F1 score revealed that among these models, the pre-trained InceptionV3 model utilizing transfer learning demonstrated the most effective performance. Moreover, these approaches provide a complete classification of unprocessed images without requiring manual extraction of attributes.