Brain tumours pose a significant global health challenge, with high mortality rates due to delayed or inaccurate diagnoses. Early detection and precise classification are crucial for effective treatment. However, traditional diagnosis relies on manual MRI interpretation, which is time-consuming and prone to human error. Radiologists often face large volumes of complex MRI data, leading to missed detections or misclassifications, particularly in small or low-contrast tumours. This study explores deep learning, specifically Convolutional Neural Networks (CNNs), for automated brain tumour detection and classification. Using extensive MRI datasets covering gliomas, meningiomas, and pituitary tumours, a Sequential CNN model achieved 99% accuracy in detecting abnormal regions. A modified VGG16 model attained 98% accuracy in tumour classification. These models outperform traditional methods by providing faster, more accurate diagnoses, reducing radiologists’ workload, and expediting clinical decisions. Beyond clinical diagnostics, CNN-based models integrated into telemedicine could revolutionise healthcare, especially in low-resource areas with limited radiologists. By ensuring consistent and accurate diagnoses, this research contributes to improved patient outcomes and more efficient medical care.