Enhanced Brain Tumor Detection with Deep Learning: A Comparative Study of CNN, VGG16 and ResNet Models

Brain tumor is one of the most serious health problems. Some brain tumors can also be cancerous and cause brain cancer. These tumors can form in different parts of the brain, like meningiomas tumors in the meninges, pituitary tumors in pituitary gland and other tumors can be identified by the type of cells they are made of, like gliomas. CNN, VGG-16, and ResNet are employed to classify and identify the type of brain tumor with the help of images in the medical field. The study highlights the significance of the precise diagnosis to cure brain tumors by using advanced image techniques like MRI (Magnetic Resonance Imaging). Deep learning algorithms have resulted in astounding outcomes in classifying and identifying brain tumors by using medical imaging applications. The research utilizes CNN, VGG-16, and ResNet to classify and identify the brain tumor from MRI scans. The study uses brain tumor dataset consisting of 4 directories having 3096 images of MRI scans acquired from Kaggle. By undergoing intensive training, the three models have illustrated favourable findings in precisely identifying the type of brain tumor. The training approach involves importing data, preprocessing the data, dividing it into training, testing and validation sets, model building and using Adam optimizer. Evaluation metrics like learning curve, which shows loss and accuracy curves, verify the efficacy of the model. In our study, by applying these models we achieved great results. CNN attained 97.4 percent accuracy, VGG-16 attained 98 percent accuracy and ResNet attained 88.81 percent accuracy. Comparing these models VGG-16 outperformed the other two models with high accuracy i.e., 98 percent accuracy which helps the medical professionals to make accurate decisions while treating the tumors.