Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Jun 20, 2025
Open Peer Review Period: Jun 20, 2025 - Aug 15, 2025
Date Accepted: Dec 23, 2025
(closed for review but you can still tweet)
Two-Minute Deep Learning-Powered Brain Quantitative Mapping: Accelerating Clinical Imaging with Synthetic MRI
ABSTRACT
Background:
Quantitative magnetic resonance imaging (qMRI) is an advanced technique that can map the physical properties of different tissues, offering crucial insights for disease diagnosis, monitoring, and assessment of treatment efficacy. Nonetheless, the practical application of this technology is indeed constrained by several factors, with the most notable being the protracted scanning duration.
Objective:
To explore using synthetic magnetic resonance imaging (synthetic MRI) with deep learning reconstruction for ultra-fast whole brain quantitative imaging and evaluate its clinical feasibility.
Methods:
We recruited 151 healthy subjects. Routine and fast synthetic MRI protocols were used in brain imaging. High-resolution quantitative maps obtained by routine scanning were used as reference images, and super-resolution generative adversarial networks (SRGAN) was used to build a super-resolution network for fast scanning images to achieve acceleration. For each quantitative map, about 2900 slices were utilized for model training, and about 700 images were reserved for test. Paired t-test, linear regression, and Bland-Altman analysis were employed to evaluate the accuracy of quantitative values under deep learning, routine scan and fast scan. P＜0.05 was statistically significant.
Results:
Compared with routine scans, fast scanning time is shortened by half. After deep learning reconstruction, the accuracy of quantitative values has not changed at different tissues and brain regions (linear regression analysis, T1: R2 = 0.98, T2: R2 = 0.97, PD: R2 = 0.99, p < 0.0001), which does not affect the quantitative diagnosis ability of typical lesions (paired t-test, T1: p value = 0.67, T2: p value = 0.73, PD: p value = 0.75). At the same time, the image quality has been greatly improved.
Conclusions:
Deep learning applications help generate quantitative metrics similar to routine scans from fast scans and ensure high-quality images. This may have a potential impact on reducing clinical scan time.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.