Kim Jong-Min, Ha Su Min
J Korean Soc Radiol. 2025 Mar;86(2):227-235. doi: 10.3348/jksr.2025.0012. Epub 2025 Mar 26.
Breast MRI is the most sensitive imaging modality for detecting breast cancer. However, its widespread use is limited by factors such as extended examination times, need for contrast agents, and susceptibility to motion artifacts. Artificial intelligence (AI) has emerged as a promising solution for these challenges by enhancing the efficiency and accuracy of breast MRI in multiple domains. AI-driven image reconstruction techniques have significantly reduced scan times while preserving image quality. This method outperforms traditional parallel imaging and compressed sensing. AI has also shown great promise for lesion classification and segmentation, with convolutional neural networks and U-Net architectures improving the differentiation between benign and malignant lesions. AI-based segmentation methods enable accurate tumor detection and characterization, thereby aiding personalized treatment planning. An AI triaging system has demonstrated the potential to streamline workflow efficiency by identifying low-suspicion cases and reducing the workload of radiologists. Another promising application is synthetic breast MR image generation, which aims to generate contrast enhanced images from non-contrast sequences, thereby improving accessibility and patient safety. Further research is required to validate AI models across diverse populations and imaging protocols. As AI continues to evolve, it is expected to play an important role in the optimization of breast MRI.
乳腺磁共振成像（Breast MRI）是检测乳腺癌最敏感的成像方式。然而，其广泛应用受到多种因素的限制，如检查时间长、需要使用造影剂以及易受运动伪影影响。人工智能（AI）通过在多个领域提高乳腺MRI的效率和准确性，已成为应对这些挑战的一种有前景的解决方案。人工智能驱动的图像重建技术在保持图像质量的同时，显著缩短了扫描时间。这种方法优于传统的并行成像和压缩感知。人工智能在病变分类和分割方面也显示出巨大潜力，卷积神经网络和U-Net架构提高了良性和恶性病变之间的区分能力。基于人工智能的分割方法能够实现准确的肿瘤检测和特征描述，从而有助于个性化治疗方案的制定。一个人工智能分流系统已证明，通过识别低可疑病例和减轻放射科医生的工作量，有潜力提高工作流程效率。另一个有前景的应用是合成乳腺MR图像生成，其目的是从非增强序列生成对比增强图像，从而提高可及性并保障患者安全。需要进一步研究以在不同人群和成像协议中验证人工智能模型。随着人工智能不断发展，预计它将在乳腺MRI的优化中发挥重要作用。
