Prediction and Classification of Brain Tumor using Modified Transfer Learning

This study uses a novel method of segmenting and computing the tissue area in brain MRI images to predict the presence of brain tumors and classify the type of tumor into common types (pituitary, meningioma, and glioma). The tissues that make up the brain have intricate boundaries. In order to more accurately classify the different types of tumors, we have thus divided the brain tissues into three segments and examined for enlarged segments using Convolutional Neural Networks (CNNs) and a Modified CNN that uses transfer learning and VGGNet. The paper's methodology involves using CNNs with Transfer Learning for efficient image segmentation and clustering, followed by VGGNet for noise reduction and image segmentation refining. We are optimizing the internal parameters of VGGNet in order to outperform the competition. The volume and density of malignant clusters are computed through pixel counting. This kind of method works pretty well for figuring out tissue volume in a wide range of illnesses that affect people of all ages.