Abstract: Brain tumours pose significant challenges in diagnosis and treatment due to their diverse appearances and complex features. Manual interpretation by radiologists often leads to subjectivity and variability in diagnoses, while traditional machine learning approaches may struggle to capture intricate patterns from medical imaging data effectively. To address these challenges, this study proposes a robust brain tumour classification system based on the ResNet50 architecture, a deep convolutional neural network known for its effectiveness in image classification tasks. Leveraging deep learning techniques, the system automates feature extraction and learns hierarchical representations directly from brain MRI scans. By initializing the ResNet50 model with pre-trained weights and fine-tuning it on a comprehensive dataset, the system achieves high classification accuracy, sensitivity, and specificity. Extensive experimentation and validation demonstrate the system's capability to accurately distinguish between gliomas, meningiomas, and pituitary tumours, providing clinicians with a reliable tool for improved diagnosis and patient care.

Keywords: Classification system, ResNet50 architecture, Deep learning Convolutional neural network, MRI scans

| DOI: 10.17148/IARJSET.2024.11585