Deep learning of personalized priors from past MRI scans enables fast, quality-enhanced point-of-care MRI with low-cost systems

2025/05/05 by Tal Oved, Oved, Tal, Beatrice Lena +11 · 2 citations
Medicine · #Advanced MRI Techniques and Applications #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #MRI in cancer diagnosis #Machine Learning (cs.LG) #Radiomics and Machine Learning in Medical Imaging #Signal Processing (eess.SP) #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2505.02470

openalex publication_date 2025/05/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Magnetic resonance imaging (MRI) offers superb-quality images, but its accessibility is limited by high costs, posing challenges for patients requiring longitudinal care. Low-field MRI provides affordable imaging with low-cost devices but is hindered by long scans and degraded image quality, including low signal-to-noise ratio (SNR) and tissue contrast. We propose a novel healthcare paradigm: using deep learning to extract personalized features from past standard high-field MRI scans and harnessing them to enable accelerated, enhanced-quality follow-up scans with low-cost systems. To overcome the SNR and contrast differences, we introduce ViT-Fuser, a feature-fusion vision transformer that learns features from past scans, e.g. those stored in standard DICOM CDs. We show that a single prior scan is sufficient, and this scan can come from various MRI vendors, field strengths, and pulse sequences. Experiments with four datasets, including glioblastoma data, low-field (50mT), and ultra-low-field (6.5mT) data, demonstrate that ViT-Fuser outperforms state-of-the-art methods, providing enhanced-quality images from accelerated low-field scans, with robustness to out-of-distribution data. Our freely available framework thus enables rapid, diagnostic-quality, low-cost imaging for wide healthcare applications.

Citations

• MR imaging in the low-field: Leveraging the power of machine learning
• A Generative Diffusion Model to Solve Inverse Problems for Robust in-NICU Neonatal MRI
• Rethinking FID: Towards a Better Evaluation Metric for Image Generation
• Repeatability of image quality in very low field MRI
• On‐site construction of a point‐of‐care low‐field MRI system in Africa
• Low‐field MRI: Clinical promise and challenges
• Physics-Driven Deep Learning for Computational Magnetic Resonance Imaging
• Physics-Driven Deep Learning for Computational Magnetic Resonance Imaging: Combining physics and machine learning for improved medical imaging
• ReconFormer: Accelerated MRI Reconstruction Using Recurrent Transformer
• Unsupervised MRI Reconstruction via Zero-Shot Learned Adversarial\n Transformers
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Joint multi-contrast Variational Network reconstruction (jVN) with\n application to rapid 2D and 3D imaging
• Low‐field MRI: An MR physics perspective
• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
• The Perception-Distortion Tradeoff
• Learning a Variational Network for Reconstruction of Accelerated MRI\n Data
• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Image quality assessment: from error visibility to structural similarity
• Random sample consensus

Cited by

• Physics-Informed Deep Unrolled Network for Portable MR Image Reconstruction
• A Tutorial on MRI Reconstruction: From Modern Methods to Clinical Implications

Related

• Accelerating Low-field MRI: From Compressed Sensing to Deep Learning Reconstruction with CNNs and Transformers
• MR imaging in the low-field: Leveraging the power of machine learning
• Deep Learning for Low-Field to High-Field MR: Image Quality Transfer with Probabilistic Decimation Simulator
• Learning Deep MRI Reconstruction Models from Scratch in Low-Data Regimes
• One for Multiple: Physics-informed Synthetic Data Boosts Generalizable Deep Learning for Fast MRI Reconstruction