Brain Tumour Detection Using Deep Learning: A CNN-Based Approach

Brain tumours are one of the biggest threats to life-threatening cancers, and timely and accurate recognition is critical for effective treatment planning and enhancing patient outcomes. Manual analysis of magnetic resonance imaging (MRI) by radiologists is standard diagnostic practice, but it is often time-consuming and can lead to inter-observer variability, leading to delayed or inaccurate diagnosis. In current investigation, we propose a folding deep learning (DL) framework for neural networks (CNNs) recorded by MRI scans of automated brain tumor detection. This model was developed using published data records containing either axis MRI images marked as Tumours or not tumor. Use preprocessing techniques such as grey level gray levels, image size, and data expansion (rotation, flipping, zoom) to improve model generalization and over-adaptation. This model is trained, verified and evaluated in a split of 80:20 train tests based on accuracy, accuracy, recall and F1 scores. The proposed model achieves accuracy of over 95% and demonstrates its effectiveness in distinguishing healthy brain tissue and tumor-related brain tissue. Furthermore, visualizations such as confusion matrix and sample predictions provide insight into model's decision process. Future research will examine the inclusion of tumor classifications of more complex architectures such as resets and efficient nets, including multiclass classifications (such as glioma, meningioma, pituitary gland), and integration into real-time diagnostic systems.