人工智能
分割
计算机科学
计算机视觉
特征(语言学)
图像分割
模式识别(心理学)
尺度空间分割
脑瘤
面子(社会学概念)
适应性
基于分割的对象分类
图像(数学)
特征提取
边界(拓扑)
医学影像学
深度学习
图像融合
作者
Yuyu Chen,Ziyi Zhang,Yufang Dong
出处
期刊:Conference proceedings
[Institute of Electrical and Electronics Engineers]
日期:2025-10-05
卷期号:: 388-393
标识
DOI:10.1109/smc58881.2025.11343591
摘要
Medical image segmentation plays a crucial role in clinical diagnosis. However, the accuracy of segmentation is still challenged by factors such as blurred target boundaries, complex shapes, and difficulties in data annotation. Although deep learning-based methods have made significant progress, these approaches still face challenges in handling issues such as blurred tumor boundaries, multi-scale distributions, and the difficulty in capturing small lesions. To address these challenges, we innovatively propose a brain tumor segmentation model based on SAM—SAM-BrainFocus. This model, built upon the original SAM approach, introduces a brain tumor feature extractor, a tumor hint generator, and a tumor boundary refiner, which significantly enhance segmentation accuracy and robustness. It has achieved breakthrough progress in multi-scale feature fusion and adaptability to complex targets. Experimental results demonstrate that SAM-BrainFocus outperforms existing mainstream models in brain tumor image segmentation tasks, achieving higher accuracy and reliability.
科研通智能强力驱动
Strongly Powered by AbleSci AI