Masked and Predictive Self-Supervised Foundation Models for 3D Brain MRI

📄 arXiv: 2606.13315 📥 PDF

作者: Esra Ergün, Hersh Chandarana, Dan Sodickson, Gözde Ünal

分类: cs.CV, eess.IV

发布日期: 2026-06-12

💡 一句话要点

提出自监督基础模型以提升3D脑MRI疾病检测效果

🎯 匹配领域: 支柱二：RL算法与架构 (RL & Architecture) 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 自监督学习 医学影像 脑MRI 疾病检测 深度学习 模型预训练 频谱域重建 方差协方差正则化

📋 核心要点

  1. 现有的MRI基础模型研究主要集中在分割和密集预测任务，缺乏对基于MRI的疾病检测的系统性探讨。
  2. 本研究提出了两种自监督预训练范式，分别为基于重建的MAE和基于预测的JEPA，旨在提升MRI疾病检测的性能。
  3. 实验结果显示，MAE结合频谱域监督在下游任务中表现优越，尤其在高频解剖结构的识别上提升显著。

📝 摘要（中文）

自监督基础模型在医学影像中展现出强大的潜力。然而，现有的MRI基础模型研究主要集中在分割和密集预测任务上，对于基于MRI的疾病检测的系统性研究仍然有限。本研究探讨了两种主要的自监督预训练范式：通过Masked Autoencoders (MAE)进行的重建学习和通过Joint Embedding Predictive Architectures (JEPA)进行的预测表示学习。我们引入了一种新颖的频谱域重建损失，以增强对细粒度解剖结构的敏感性，并在JEPA框架中整合了方差-协方差正则化（VCR），以鼓励去相关的潜在表示。实验结果表明，自监督目标的设计对医学基础模型的预训练至关重要，且每个目标的下游效益与任务结构的相关性密切相关。

🔬 方法详解

问题定义：本研究旨在解决现有MRI基础模型在疾病检测任务中的不足，尤其是缺乏系统性自监督学习的探讨。

核心思路：通过引入MAE和JEPA两种自监督学习方法，结合新颖的损失函数设计，提升模型对细粒度解剖结构的敏感性和潜在表示的去相关性。

技术框架：整体架构包括两个主要模块：MAE用于重建学习，JEPA用于预测表示学习。模型在异构单对比MRI体积上进行预训练，且不进行模态拼接。

关键创新：引入频谱域重建损失和方差-协方差正则化（VCR）作为辅助目标，显著提升了模型在下游疾病检测任务中的表现。

关键设计：MAE采用频谱域监督，强调对高频解剖结构的敏感性；JEPA则通过VCR鼓励潜在表示的去相关性，确保模型在多维特征空间中的有效性。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明，结合频谱域监督的MAE在五个下游疾病检测任务中表现优越，相较于基线模型，性能提升幅度达到了显著的水平，尤其在高频解剖结构的识别上表现突出。

🎯 应用场景

该研究的潜在应用领域包括医学影像分析、疾病早期检测和个性化医疗。通过提升MRI疾病检测的准确性，能够为临床决策提供更可靠的支持，未来可能推动相关技术在医疗领域的广泛应用。

📄 摘要（原文）

Self-supervised foundation models have shown strong promise in medical imaging. However, existing MRI foundation-model studies have primarily emphasized segmentation and dense prediction tasks, while systematic investigation of self-supervised foundation models for MRI-based disease detection remains limited. In this work, we investigate two major self-supervised pretraining paradigms for MRI-based disease detection: reconstruction-based learning via Masked Autoencoders (MAE) and predictive representation learning via Joint Embedding Predictive Architectures (JEPA). We study the role of auxiliary objectives by introducing a novel spectral-domain reconstruction loss for MAE to enhance sensitivity to fine-grained anatomical structure, and by integrating variance--covariance regularization (VCR) within our JEPA framework to encourage decorrelated latent representations. Our models are pretrained on heterogeneous single-contrast MRI volumes in a contrast-agnostic setting, without modality concatenation. Across five downstream disease detection tasks, our results highlight the importance of self-supervised objective design for medical foundation model pretraining, demonstrating that the downstream benefit of each objective is determined by its relevance to the task's structure. Specifically, spectral regularization yields the largest improvements when the downstream discriminative signal is characterized by strong high-frequency anatomical structures, while covariance regularization is most beneficial when discriminative information spans multiple decorrelated feature dimensions. MAE with spectral-domain supervision consistently achieves superior downstream performance for MRI-based disease detection. These findings suggest that self-supervised objectives in medical imaging encode specific biases, and their downstream benefit is fundamentally conditioned on the task's structure.