<![CDATA[Detecting brain tumours from MRI images remains a difficult challenge due to differences in tumour appearance and the necessity for high diagnostic precision. This study looks at three deep learning algorithms with varying levels of complexity: CNN as a baseline classification model, Faster R-CNN as a region-based detection method, and Mask R-CNN, which combines detection with segmentation. The dataset is divided into four categories: glioma, meningioma, pituitary, and non-tumor. The experimental results show that more advanced structures tend to perform better. The CNN model achieves an accuracy of 0.8900000000 with an F1-score of 0.8871239227, although it has problems in capturing specific tumour characteristics. Faster R-CNN enhances detection capability, with an F1-score of 0.9053533622 and an accuracy of 0.9068750000, especially when recognising tumour locations more precisely. Mask R-CNN achieves the best performance, with an accuracy of 0.9300000000 and an F1-score of 0.9288687706, indicating more consistent results across all classes. Mask R-CNN has the advantage of capturing both object position and structural features via segmentation, hence minimising misclassification. These results imply that integrating detection and segmentation is critical for improving medical image analysis. As a result, Mask R-CNN provides a more reliable method for detecting brain tumours using MRI data.]]>
Copyrights © 2026
