Automated Brain Tumor Detection Using Enhanced ResNet-50 Deep Learning Model

Brain tumors are the presence and development of malignant or non-cancerous cells in the brain, which can severely impair cognitive functionality of a person ranging from distorted memory, lack of concentration and problem solving ability. Traditional methods for brain tumor classification using MRI images rely on manual inspection or conventional Convolutional Neural Networks (CNN's), which often struggle with interpretability and generalization due to the complexity and variability in medical images. The purpose of this research is to improve the accuracy and robustness of brain tumor classification by leveraging the ResNet50 architecture, a pre-trained deep convolutional neural network (CNN), combined with advanced image preprocessing techniques. The methodology involves contour-based cropping to isolate the tumor region from irrelevant background noise, thereby enhancing the focus on tumor-specific features. The proposed method achieves 98.33% accuracy, outperforming existing approaches. The novelty of this approach lies in the combination of targeted tumor cropping, transfer learning, and augmentation, which collectively enhance the model's ability to generalize and accurately detect brain tumors in MRI images.