New Shared LoRA Framework Enhances MRI Reconstruction Efficiency

Researchers have developed a novel parameter-efficient framework called Shared LoRA for magnetic resonance imaging (MRI) reconstruction. This method freezes a pretrained SHFormer backbone and trains a single set of LoRA adapters along with a gating network. By generating undersampled inputs across various acceleration factors during training, the shared adapters learn to reconstruct images effectively, while a gating network dynamically adjusts adapter strengths based on the given acceleration factor. Experiments demonstrate that Shared LoRA achieves competitive performance with significantly fewer trainable parameters, showing stable generalization to unseen factors. AI

IMPACT This method could lead to more efficient and cost-effective MRI reconstruction techniques by reducing the need for factor-specific models.

RANK_REASON The item is an academic paper detailing a new method for MRI reconstruction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• LoRA+
• magnetic resonance imaging
• ScienceCast
• SHFormer

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →