🧠 AI nails brain tumor classification with new deep learning method
Researchers introduced an adaptive hierarchical class-aware deep ensemble strategy for better brain tumor classification using MRI. The study involved experts VGG19 and Darknet53, tested on 7,200 stratified images across glioma, meningioma, no tumor, and pituitary classes. On 1,080 unseen test images, the proposed strategy achieved 98.15% accuracy and 98.14% macro F1-score. This method uses a router to consult an expert map during model disagreements for robust decisions. The approach offers a transparent and auditable decision pathway for future medical applications. 🧠