Accelerated multi-modal magnetic resonance (MR) imaging is a new and effective solution for fast MR imaging, providing superior performance in restoring the target modality from its undersampled counterpart with guidance an auxiliary modality. However, existing works simply combine as prior information, lacking in-depth investigations on potential mechanisms fusing different modalities. Further, they usually rely convolutional neural networks (CNNs), which limited by intrinsic locality...

Federated learning (FL) can be used to improve data privacy and efficiency in magnetic resonance (MR) image reconstruction by enabling multiple institutions collaborate without needing aggregate local data. However, the domain shift caused different MR imaging protocols substantially degrade performance of FL models. Recent techniques tend solve this enhancing generalization global model, but they ignore domain-specific features, which may contain important information about device...

Super-resolving the magnetic resonance (MR) image of a target contrast under guidance corresponding auxiliary contrast, which provides additional anatomical information, is new and effective solution for fast MR imaging. However, current multi-contrast super-resolution (SR) methods tend to concatenate different contrasts directly, ignoring their relationships in clues, e.g., high-and low-intensity regions. In this study, we propose separable attention network (comprising high-intensity...

Federated learning enables multiple hospitals to cooperatively learn a shared model without privacy disclosure. Existing methods often take common assumption that the data from different have same modalities. However, such setting is difficult fully satisfy in practical applications, since imaging guidelines may be between hospitals, which makes number of individuals with set modalities limited. To this end, we formulate practical-yet-challenging cross-modal vertical federated task, small...

Super-resolving the Magnetic Resonance (MR) image of a target contrast under guidance corresponding auxiliary contrast, which provides additional anatomical information, is new and effective solution for fast MR imaging. However, current multi-contrast super-resolution (SR) methods tend to concatenate different contrasts directly, ignoring their relationships in clues, e.g., high-intensity low-intensity regions. In this study, we propose separable attention network (comprising priority...

Federated learning (FL) facilitates collaborative among multiple clients in a distributed manner, while ensuring privacy protection. However, its performance is inevitably degraded as suffering data heterogeneity, i.e., non-IID data. In this paper, we focus on the feature distribution skewed FL scenario, which widespread real-world applications. The main challenge lies shift caused by different underlying distributions of local datasets. While previous attempts achieved progress, few studies...

Federated learning (FL) can be used to improve data privacy and efficiency in magnetic resonance (MR) image reconstruction by enabling multiple institutions collaborate without needing aggregate local data. However, the domain shift caused different MR imaging protocols substantially degrade performance of FL models. Recent techniques tend solve this enhancing generalization global model, but they ignore domain-specific features, which may contain important information about device...

Federated Learning (FL) enables multiple clients to collaboratively learn in a distributed way, allowing for privacy protection. However, the real-world non-IID data will lead client drift which degrades performance of FL. Interestingly, we find that difference logits between local and global models increases as model is continuously updated, thus seriously deteriorating FL performance. This mainly due catastrophic forgetting caused by heterogeneity clients. To alleviate this problem,...

The core problem of Magnetic Resonance Imaging (MRI) is the trade off between acceleration and image quality. Image reconstruction super-resolution are two crucial techniques in (MRI). Current methods designed to perform these tasks separately, ignoring correlations them. In this work, we propose an end-to-end task transformer network (T$^2$Net) for joint MRI super-resolution, which allows representations feature transmission be shared multiple achieve higher-quality, super-resolved...

Accelerated multi-modal magnetic resonance (MR) imaging is a new and effective solution for fast MR imaging, providing superior performance in restoring the target modality from its undersampled counterpart with guidance an auxiliary modality. However, existing works simply combine as prior information, lacking in-depth investigations on potential mechanisms fusing different modalities. Further, they usually rely convolutional neural networks (CNNs), which limited by intrinsic locality...

Federated learning enables multiple hospitals to cooperatively learn a shared model without privacy disclosure. Existing methods often take common assumption that the data from different have same modalities. However, such setting is difficult fully satisfy in practical applications, since imaging guidelines may be between hospitals, which makes number of individuals with set modalities limited. To this end, we formulate practical-yet-challenging cross-modal vertical federated task, shape...

While multi-modal learning has been widely used for MRI reconstruction, it relies on paired data which is difficult to acquire in real clinical scenarios. Especially the federated setting, common situation that several medical institutions only have single-modal data, termed modality missing issue. Therefore, infeasible deploy a standard framework such conditions. In this paper, we propose novel communication-efficient framework, namely Fed-PMG, address challenge reconstruction....