医学
脑瘤
基础(证据)
核医学
计算机科学
脑病
放射科
磁共振成像
医学影像学
医学物理学
作者
Kaiqi Qiang,Hongming Xu,Qibin Zhang,Wu Chen,Fengyu Cong,Haozhe Piao
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-04-08
卷期号:: 1-4
被引量:1
标识
DOI:10.1109/isbi61048.2026.11515680
摘要
Magnetic resonance imaging (MRI) plays a critical role in brain tumor diagnosis and segmentation. Recent medical foundation models have demonstrated strong performance in full-modality brain tumor segmentation by jointly utilizing multiple MRI sequences. However, in clinical practice, missing modalities caused by patient motion, contrast agent intolerance, or acquisition constraints often lead to degraded segmentation accuracy. To address this issue, we propose TSFMD, a two-stage framework for brain Tumor Segmentation under missing-modality conditions via Foundation Model Distillation. In the first stage, the student model is pre-trained under the guidance of BrainSegFounder using a Centered Kernel Alignment (CKA)-based distillation strategy. In the second stage, the student model is finetuned for tumor segmentation. Extensive experiments on the BraTS2020 dataset demonstrate that our method achieves superior segmentation performance under missing-modality scenarios and outperforms state-of-the-art (SOTA) methods in inference efficiency.
科研通智能强力驱动
Strongly Powered by AbleSci AI