Deep learning for multimodal brain tumor segmentation: Architectures, fusion, robust learning, and deployment perspectives

2026-08-10

PRE

Yi Zhou

Jeevan Kanesan *

Chee-Onn Chow

Chengbao Xu

DOI：10.1016/j.compmedimag.2026.102807

Original

Original request for help

Save

Abstract

En 中文

• This survey reviews multimodal brain tumor segmentation from a failure-oriented deployment perspective. • It links input reliability, fusion–architecture co-design, subregion failure, and clinical triage. • It analyzes CNNs, Transformers, state-space models, diffusion methods, and foundation models. • It highlights persistent ET/TC fragility under missing, degraded, and shifted MRI conditions. • It emphasizes stress testing, uncertainty calibration, and human-in-the-loop clinical usability.

Deep learning for multimodal brain tumor segmentation: Architectures, fusion, robust learning, and deployment perspectives

Abstract

Journal

Organization

Cited Papers

Citing Papers