13-May-2026
深度学习驱动的MRI扫描流程智能化探索
ResearchPeer-Reviewed Publication
随着人工智能(AI)在医学影像领域的快速发展,深度学习算法已在影像识别、结构分割和辅助诊断中展现出巨大潜力。然而,目前大多数AI系统仍主要应用于影像获取后的离线分析阶段,在MRI扫描过程中对图像采集与定位的实时辅助能力仍十分有限。近期,研究团队提出 Lumbar VNet Pro(LVP)系统,率先探索将深度学习算法直接嵌入MRI扫描设备工作流程,实现实时定位、自动结构分割与扫描参数优化。该系统基于2453例MRI数据训练,并在多中心1522例患者中开展临床验证。结果显示,该系统在结构定位(Dice=0.93)、分割(Dice=0.92)以及椎间盘识别等任务中均表现出优异性能,同时显著提高MRI定位精度并降低观察者间差异。该研究表明,将AI直接集成到影像设备中,可构建“采集—分析—反馈”的闭环影像流程,为未来智能化医学影像系统提供新的技术路径。相关成果以 “Clinical Application of Deep Learning for Spine MRI Interpretation” 为题发表在Research上。
- Journal
- Research
- Funder
- National Natural Science Foundation of China, Guangdong Provincial People’s Hospital Full-time High-level Talent Introduction Foundation, Science and Technology Program of Guangzhou, Science and Technology Program of Guangzhou—Key Project, Guangdong Natural Science Foundation, Scientific Research Fund of the Medical Equipment Society of Guangdong Province, Guangdong Basic and Applied Basic Research Foundation, Presidential Foundation of Zhujiang Hospital, Southern Medical University, Clinical Medicine Special Fund Project of the Zhejiang Medical Association, President’s fund of Nanfang Hospital, Science and Technology Projects of Guangzhou, Provincial Training Program of Innovation and Entrepreneurship for Undergraduates