Brain Tumor Detection Using Transfer Learning Based CNN with ResNet50

Abstract

Brain tumor detection is a critical task in medical image analysis, where accurate and timely identification can support effective clinical decision-making. Manual examination of Magnetic Resonance Imaging (MRI) scans is often complex, time-consuming, and susceptible to variations in interpretation. This study presents a deep learning-based approach for brain tumor classification using transfer learning with a Convolutional Neural Network (CNN) based on the ResNet50 architecture. The proposed approach utilizes pretrained features to improve classification performance while reducing the dependency on large-scale training datasets. To enhance model efficiency and generalization, image preprocessing techniques including resizing, normalization, and data augmentation are incorporated into the training pipeline. The study also integrates perspectives inspired by Indian Knowledge Systems (IKS), particularly structured reasoning and pattern recognition principles, which are conceptually aligned with the hierarchical feature-learning process of deep neural networks. This integration provides an additional perspective for interpreting the progressive learning and classification process of the model. The proposed approach demonstrates reliable performance in distinguishing tumor and non-tumor MRI images and has the potential to assist medical professionals in supporting accurate and timely decision-making. The findings highlight the applicability of transfer learning and structured computational approaches for intelligent medical image analysis.