Low field MRI scans reveal limited individual brain detail compared to high field scans
You Cannot Recover What Was Never Measured: Quantifying the Information Ceiling of Ultra-Low-Field MRI Super-Resolution
Computer Vision and Pattern RecognitionMachine Learning
Summary
MRI machines with lower strength magnets produce images that look blurrier and contain less detail about a person’s brain. The authors measured how much real low-strength MRI scans actually capture unique brain features and found they lose small-scale details that stronger machines show. They tested advanced AI models designed to enhance these scans and found that these models often create details that aren’t really there, making the enhanced images look more like guesses than true recoveries. This work helps clarify the limits of what can be trusted in low-strength MRI enhancements.
What this means in practice
- •For medical imaging developers: Establish guidelines for designing and evaluating super-resolution models that reflect true recoverable brain details from low-field MRI data.
- •For radiology equipment teams: Avoid overestimating the information content in low-field MRI scans when improving portable MRI hardware or software for clinical use.
Authors
Prathamesh Pradeep Khole, Shreya Handa, Utkarsh Gupta, Razvan Marinescu
Abstract
Generative super-resolution models can turn portable 64 mT MRI into images that look like 3T scans, and the field evaluates them with PSNR, SSIM, and pixelwise uncertainty, most often on pairs built by synthetically degrading high-field images. Prior work acknowledges that these models hallucinate and that the problem is ill posed, but to our knowledge no study measures how much information about the individual subject the real low-field scan actually contains. We measure it. Using paired 64 mT and 3T scans of the same subjects from three public datasets, and a measurement protocol validated on tests whose correct answer is known in advance, we find that, judged over the whole brain, real 64 mT scans carry structure specific to the individual only down to approximately 3 to 4 mm half-pitch in plane, and coarser still through plane. Standard synthetic degradations preserve subject information roughly 1 mm beyond this ceiling, so models trained and benchmarked on synthetic pairs are evaluated on information that real scanners never record. We then test trained diffusion models and a publicly released external model on real paired acquisitions; 24 trained runs of five architectures (GAN, diffusion, and transformer families) give the coverage of the audit. On every subject where faithfulness can be measured, fine output detail is no more correlated with the subject's own 3T scan than with a stranger's, while sample-variance uncertainty does not distinguish fabricated structure from reconstruction difficulty. Because PSNR and SSIM score resemblance to a reference rather than whether detail belongs to the subject, a benchmark scored by them cannot tell recovery from fabrication. Code for the measurement protocol will be released so that recoverability claims can be tested for newer models.