DARE-FUSE: domain aligned evidence guided learning for joint brain tumor MRI segmentation and classification

1

DARE-FUSE is a unified framework designed for MRI segmentation and classification of brain tumors under limited samples and labels.
2

The framework employs dual encoders and a Domain Alignment Refiner to create task-aligned representations for segmentation and classification.
3

DARE-FUSE utilizes U-SEG for feature decoding and SEGU for pixel-wise uncertainty to enhance segmentation accuracy.
4

The classification branch includes CPG for predictions and multi-scale Grad-CAM++ for evidence generation.
5

DARE-FUSE demonstrates superior performance on BraTS benchmarks and supports interpretable decision-making in clinical settings.

Npj Digital Medicine

by Yuqi Liu, Chen Sun, Yuning Niu, Xu Wang, Zehua Yue, Tieqiang Zhang, Jiang Li, Xiudong Guan, Dainan Zhang, Wang Jia
February 2, 2026

1