Technology Category
Machine Learning: Unsupervised & Self-Supervised LearningComputer Vision: Multi-modal VisionKnowledge Representation and Reasoning: Qualitative Reasoning
Self-Supervised Weighted Image Guided Quantitative MRI Super-Resolution
🤖 AI Summary
High-resolution quantitative MRI (qMRI) remains clinically limited due to prohibitively long acquisition times. To address this, we propose a physics-informed self-supervised super-resolution framework that eliminates the need for high-resolution qMRI ground-truth labels. Leveraging conventional high-resolution weighted MRI (wMRI) as a physical prior, our method jointly enforces MR signal model consistency and k-space degradation fidelity within a Bayesian maximum-a-posteriori (MAP) estimation framework, enabling end-to-end learning. This work introduces the first wMRI-guided self-supervised qMRI super-resolution paradigm, fully decoupling training from HR qMRI annotations and supporting cross-sequence generalization and multi-center, scanner-agnostic deployment. Given only a 1-minute low-resolution qMRI input, our method reconstructs high-fidelity T1/T2 maps comparable in quality to those acquired in 5 minutes. Validation on independent in vivo data confirms robust performance, substantially enhancing the clinical feasibility of qMRI.
Technology Category
Application Category
📝 Abstract
High-resolution (HR) quantitative MRI (qMRI) relaxometry provides objective tissue characterization but remains clinically underutilized due to lengthy acquisition times. We propose a physics-informed, self-supervised framework for qMRI super-resolution that uses routinely acquired HR weighted MRI (wMRI) scans as guidance, thus, removing the necessity for HR qMRI ground truth during training. We formulate super-resolution as Bayesian maximum a posteriori inference, minimizing two discrepancies: (1) between HR images synthesized from super-resolved qMRI maps and acquired wMRI guides via forward signal models, and (2) between acquired LR qMRI and downsampled predictions. This physics-informed objective allows the models to learn from clinical wMRI without HR qMRI supervision. To validate the concept, we generate training data by synthesizing wMRI guides from HR qMRI using signal equations, then degrading qMRI resolution via k-space truncation. A deep neural network learns the super-resolution mapping. Ablation experiments demonstrate that T1-weighted images primarily enhance T1 maps, T2-weighted images improve T2 maps, and combined guidance optimally enhances all parameters simultaneously. Validation on independently acquired in-vivo data from a different qMRI sequence confirms cross-qMRI sequence generalizability. Models trained on synthetic data can produce super-resolved maps from a 1-minute acquisition with quality comparable to a 5-minute reference scan, leveraging the scanner-independent nature of relaxometry parameters. By decoupling training from HR qMRI requirement, our framework enables fast qMRI acquisitions enhanced via routine clinical images, offering a practical pathway for integrating quantitative relaxometry into clinical workflows with acceptable additional scan time.
Problem
Research questions and friction points this paper is trying to address.
Enhances low-resolution quantitative MRI using high-resolution weighted MRI guidance.
Eliminates need for high-resolution ground truth in training via self-supervised learning.
Reduces acquisition time for quantitative MRI relaxometry in clinical settings.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Self-supervised framework uses weighted MRI as guidance without high-resolution ground truth
Physics-informed Bayesian inference minimizes discrepancies between synthesized and acquired images
Deep neural network learns super-resolution mapping from synthetic data for fast clinical scans
🔎 Similar Papers
2024-10-05arXiv.orgCitations: 0
2023-12-18arXiv.orgCitations: 7
