Generalizable synthetic MRI with physics-informed convolutional networks

TL;DR

The paper presents a physics-informed GAN framework to synthesize four standard brain MRI contrasts from a single five-minute MR-STAT acquisition by first predicting effective q*-maps and then applying a signal model. This approach aims to generalize to unseen contrasts, addressing limitations of end-to-end synthMRI methods. On a 55-subject dataset, the physics-informed method matches end-to-end performance for PDw, T1w, T2w, and T2-FLAIR (SSIM ~0.75 and PSNR ~22–23 dB) and demonstrates retrospective contrast adjustment for unseen sequences (T12w, TI400, DIR), with CNR values close to ground truth for some contrasts and notable artifacts for others. The method offers interpretability, reduced scan time, and potential for flexible, post hoc contrast selection, advancing clinically feasible synthMRI for accelerated neuroimaging.

Abstract

In this study, we develop a physics-informed deep learning-based method to synthesize multiple brain magnetic resonance imaging (MRI) contrasts from a single five-minute acquisition and investigate its ability to generalize to arbitrary contrasts to accelerate neuroimaging protocols. A dataset of fifty-five subjects acquired with a standard MRI protocol and a five-minute transient-state sequence was used to develop a physics-informed deep learning-based method. The model, based on a generative adversarial network, maps data acquired from the five-minute scan to "effective" quantitative parameter maps, here named q*-maps, by using its generated PD, T1, and T2 values in a signal model to synthesize four standard contrasts (proton density-weighted, T1-weighted, T2-weighted, and T2-weighted fluid-attenuated inversion recovery), from which losses are computed. The q*-maps are compared to literature values and the synthetic contrasts are compared to an end-to-end deep learning-based method proposed by literature. The generalizability of the proposed method is investigated for five volunteers by synthesizing three non-standard contrasts unseen during training and comparing these to respective ground truth acquisitions via contrast-to-noise ratio and quantitative assessment. The physics-informed method was able to match the high-quality synthMRI of the end-to-end method for the four standard contrasts, with mean \pm standard deviation structural similarity metrics above 0.75 \pm 0.08 and peak signal-to-noise ratios above 22.4 \pm 1.9 and 22.6 \pm 2.1. Additionally, the physics-informed method provided retrospective contrast adjustment, with visually similar signal contrast and comparable contrast-to-noise ratios to the ground truth acquisitions for three sequences unused for model training, demonstrating its generalizability and potential application to accelerate neuroimaging protocols.

Figures (7)

• Figure 1: Schematic representation of possible synthMRI approaches. The standard synthMRI approach (solid black arrows) starts with qMRI reconstruction, from which synthMRI is obtained via a signal model. End-to-end DL-based methods (dotted red arrow) have been proposed to skip the qMRI reconstruction and signal model. Our proposed physics-informed method (striped blue arrows) aims to address the lack of generalizability of the end-to-end approach by outputting effective q-maps (q*-maps) and feeding these to the signal model to obtain synthMRI.
• Figure 2: Schematic representation of the synthMRI methods. The five complex FFT images ($\textbf{x}$) were fed into the two architectures to output four contrasts ($\hat{\textbf{y}}$) directly (end-to-end) or via q*-maps followed by the physics model M (physics-informed).
• Figure 3: Q*-maps for a volunteer. A transverse slice example of the q*-maps for a volunteer. The GM/WM ratio is compared to Ref. Hagiwara2019LinearityControls and the T1 and T2 values to the median value in Ref. Bojorquez2017WhatT. The standard deviations of the PD ratio and T1 and T2 values describe inter-subject and intra-subject variability, respectively. The PD q*-maps and MR-STAT maps are normalized to the 75th percentile of the CSF volume for visualization purposes. The superior sagittal sinus is highlighted in the PD maps.
• Figure 4: Quantitative comparison of the end-to-end and physics-informed approach. The SSIM and PSNR of the contrasts synthesized using the end-to-end (blue) and physics-informed (red) approach are visualized using boxplots and statistically compared. *($0.01 < p \leq 0.05$) **($0.001 < p \leq 0.01$) ***($p \leq 0.001$) n.s. = not significant ($p > 0.05$).
• Figure 5: SynthMRI on multiple sclerosis. Contrasts are shown for conventional acquisition (top row), end-to-end (middle row), and physics-informed (bottom row) approaches, where the right columns zooming in on patches of the T2-FLAIR contrast with red and blue arrows highlighting missed and hypointense/blurred lesions, respectively.