Innovative Deep Learning Strategies for Early Detection of Brain Tumours in MRI Scans with a Modified ResNet50V2 Approach

Classifying brain tumors is vital to medical diagnosis since early identification and the distinction between benign and malignant tumors can greatly enhance patient outcomes. In this work, we use magnetic resonance imaging (MRI) scans to categorize brain cancers into four groups: pituitary, meningioma, glioma, and no tumor. Our model was trained and evaluated using a dataset of 18230 MRI pictures of the human brain. To optimize feature extraction and model performance, we suggest a modified ResNet50V2 architecture that is improved with numerous Squeeze-and-Excitation (SE) modules. The model attained a stunning 99.97% training accuracy and 97.98% validation accuracy. These findings show how our method may help improve the precision and consistency of brain tumor diagnosis, which would be a significant advancement for neurology and medical research.

Hali tarjima qilinmagan