一次评估，任意工作点：基于超网络摊销MeanFlow的3D MRI重建
One Evaluation, Any Operating Point: Hypernetwork-Amortized MeanFlow for 3D MRI Reconstruction
AI总结：
本研究提出将工作点作为输入的3D MeanFlow重建方法，通过超网络摊销数据一致性权重、加速倍数和采样模式，单次评估即可超越扩散先验，并支持自监督训练。
AI中文摘要：
生成式先验能够很好地重建加速3D MRI，但在部署时代价高昂：每个体素需要数十次网络评估，且需针对协议进行超参数调整。第三个隐藏成本是扫描仪固定的采样模式。我们将整个工作点视为输入。一个3D MeanFlow补丁网络（一种单步流模型）通过热启动的五次迭代可微共轭梯度投影进行端到端微调。一个小型超网络将工作点（数据一致性权重、加速倍数及笛卡尔采样模式本身）映射到网络的逐通道调制。有三项发现：（i）学习采集模式比任何其他工作点更有价值：在临床膝关节数据上，学习到的掩模相比协议的可变密度掩模增益高达+2.34 dB。此增益需要求解器：使用前馈重建器时，同一学习掩模在4倍加速下反而损失-1.7 dB，但结合数据一致性投影后增益+4.6 dB。（ii）单次评估非常有效：在脑部和膝关节数据上分别比20步补丁扩散先验高出+3.1 dB和+2.9 dB。三到五次评估可将前沿扩展到+6 dB，同时仅使用先验网络调用次数的四分之一。（iii）全采样目标并非必需：在采集样本的划分上进行自监督训练，重建器在4倍加速的真实数据上与其监督训练的孪生模型表现相当。最后，我们报告了失败之处及原因：基于测量能量的受试者自适应采集、将自监督与学习采集相结合，以及摊销数据一致性权重。
英文摘要：
Generative priors reconstruct accelerated 3D MRI well but pay heavily at deployment: tens of network evaluations per volume, and protocol-specific hyperparameter tuning. A third hidden cost is the scanner's fixed sampling pattern. We treat the whole operating point as an input. A 3D MeanFlow patch network (a one-step flow model) is fine-tuned end-to-end through a warm-started, five-iteration differentiable conjugate-gradient projection. A small hypernetwork maps the operating point (data-consistency weight, acceleration, and the Cartesian sampling pattern itself) to the network's per-channel modulation. Three findings follow. (i) Learning the acquisition is worth more than any other operating point: on clinical knee data, the learned mask gains up to +2.34 dB over the protocol's variable-density mask. This gain requires the solver: with a feed-forward reconstructor the same learned mask hurts at 4x (-1.7 dB), but with the data-consistency projection it adds +4.6 dB. (ii) One evaluation is highly effective: it beats a 20-step patch-diffusion prior by up to +3.1 dB on brain and +2.9 dB on knee. Three to five evaluations extend the front to +6 dB while using a quarter of the prior's network calls. (iii) Fully sampled targets are optional: trained self-supervised on a split of acquired samples, the reconstructor matches its supervised twin at 4x on real data. Finally, we report what failed and why: subject-adaptive acquisition from measured energy, combining self-supervision with learned acquisition, and amortising the data-consistency weight.
