ABSTRACT
Magnetic resonance imaging (MRI) can help visualize brain tumors (BTs). These scans reveal soft tissues details using a magnetic field and radio waves. "Brain tumor Augmented Transfer learning Categorization" (BrATCat) employs a classifier after augmenting the original images to speed up the separation of MRI scans for BT images. Most BT datasets are small and highly unbalanced. Contrasting to the highly advanced methodologies to assist in healing BTs. Hence, these facts can negatively affect the output due to learning models' underfitting and/or over-fitting. The cycleGAN increases images to image translation after the Dataset undergoes a unique 3-layered preprocessing stage. Following image augmentation, a comparative analysis of different classifiers with pre-trained models like VGG, MobileNetV2, and DenseNet201 and without them, using CNN classifier is implemented. The best model is studied and described in detail. After implementing the layered preprocessing architecture, an increase in the separation results after augmenting the predefined Dataset with synthesized images happens.
