标题： 用于神经影像分析的领域随机深度学习 显示英文标题

标题： Domain-randomized deep learning for neuroimage analysis

摘要： 深度学习通过提供前所未有的速度和准确性，彻底改变了神经影像分析。 然而，许多训练数据集的狭窄范围限制了模型的鲁棒性和泛化能力。 这一挑战在磁共振成像（MRI）中尤为严重，因为图像外观在脉冲序列和扫描仪硬件之间变化很大。 一种最近的领域随机化策略通过在具有随机强度和解剖内容的合成图像上训练深度神经网络来解决泛化问题。 通过从解剖分割图生成多样化数据，该方法使模型能够准确处理训练期间未见过的图像类型，而无需重新训练或微调。 它在包括MRI、计算机断层扫描、正电子发射断层扫描和光学相干断层扫描在内的多种模态中展示了有效性，而且在神经影像学之外的超声、电子显微镜和荧光显微镜以及X射线显微断层扫描中也有效。 本文教程论文回顾了合成驱动训练范式的原理、实现和潜力。 它强调了关键优势，如改进的泛化能力和对过拟合的抵抗力，同时讨论了计算需求增加等权衡。 最后，文章探讨了采用该技术的实际考虑因素，旨在加速开发可泛化的工具，使深度学习对没有大量计算资源或机器学习知识的领域专家更加易于使用。 显示英文摘要

摘要： Deep learning has revolutionized neuroimage analysis by delivering unprecedented speed and accuracy. However, the narrow scope of many training datasets constrains model robustness and generalizability. This challenge is particularly acute in magnetic resonance imaging (MRI), where image appearance varies widely across pulse sequences and scanner hardware. A recent domain-randomization strategy addresses the generalization problem by training deep neural networks on synthetic images with randomized intensities and anatomical content. By generating diverse data from anatomical segmentation maps, the approach enables models to accurately process image types unseen during training, without retraining or fine-tuning. It has demonstrated effectiveness across modalities including MRI, computed tomography, positron emission tomography, and optical coherence tomography, as well as beyond neuroimaging in ultrasound, electron and fluorescence microscopy, and X-ray microtomography. This tutorial paper reviews the principles, implementation, and potential of the synthesis-driven training paradigm. It highlights key benefits, such as improved generalization and resistance to overfitting, while discussing trade-offs such as increased computational demands. Finally, the article explores practical considerations for adopting the technique, aiming to accelerate the development of generalizable tools that make deep learning more accessible to domain experts without extensive computational resources or machine learning knowledge.