Joint Total Variation ESTATICS for Robust Multi-Parameter Mapping

Yaël Balbastre, Mikael Brudfors, Michela Azzarito, Christian Lambert, Martina F. Callaghan, John Ashburner

Abstract

Quantitative magnetic resonance imaging (qMRI) derives tissue-specific parameters—such as the apparent transverse relaxation rate $R_2^*$, the longitudinal relaxation rate $R_1$, and the magnetisation transfer saturation—that can be compared across sites and scanners and carry important information about the underlying microstructure. The multi-parameter mapping (MPM) protocol takes advantage of multi-echo acquisitions with variable flip angles to extract these parameters in a clinically acceptable scan time. In this context, ESTATICS performs a joint log-linear fit of multiple echo series to extract $R_2^*$ and multiple extrapolated intercepts, thereby improving robustness to motion and decreasing the variance of the estimators. In this paper, we extend this model in two ways: (1) by introducing a joint total variation (JTV) prior on the intercepts and decay, and (2) by deriving a nonlinear maximum a posteriori estimate. We evaluated the proposed algorithm by predicting left-out echoes in a rich single-subject dataset. In this validation, we outperformed other state-of-the-art methods and additionally showed that the proposed approach greatly reduces the variance of the estimated maps, without introducing bias.

Introduction

Standard clinical MRI produces "weighted" images where brightness is relative, making it difficult to compare scans across different hospitals or time points. Quantitative MRI (qMRI) addresses this by measuring absolute physical constants of tissue, such as $R_1$ (longitudinal relaxation) and $R_2^*$ (transverse relaxation), which correlate directly with biological features like myelin and iron content.

However, obtaining these measurements at high resolution (e.g., 0.8mm) significantly drops the signal-to-noise ratio (SNR), resulting in noisy or grainy maps. While researchers often use smoothing to mitigate this, traditional smoothing destroys fine details and edges. Furthermore, existing methods typically treat different image contrasts (like T1-weighted vs. PD-weighted) as independent, even though they share the same underlying anatomy and structural boundaries.

Methodology

This work builds upon the ESTATICS framework, which performs a joint fit of multiple echoes to estimate $R_2^*$ and signal intercepts. The authors introduce two key innovations:

  1. Joint Total Variation (JTV) Prior: Instead of denoising each image separately, JTV analyzes all image contrasts simultaneously. It assumes that structural edges are shared across contrasts; thus, information from a high-signal image can help denoise a lower-signal one without blurring boundaries.
  2. Nonlinear Maximum A Posteriori (MAP) Estimation: Rather than a simple log-linear fit, which can be biased by noise, the authors use a full generative model. The problem is treated as a nonlinear optimization, utilizing Fisher’s Scoring and Quadratic Bounds to ensure efficient convergence to a stable solution.

Notably, the method incorporates head motion directly into the signal reconstruction process, avoiding the data corruption often caused by standard image "reslicing."

Experimental Setup

The approach was validated using a "rich" single-subject dataset:

  • MPM Protocol: A healthy participant underwent five repeat scans in one session using a 0.8 mm Multi-Parameter Mapping protocol.
  • Validation Strategy: A Leave-One-Echo-Out cross-validation was used. The model was trained on all but one echo, and its performance was measured by how accurately it could "predict" the missing data using Rice log-likelihood scoring.

Results and Discussion

Visual comparison of denoising methods (left) and quantitative predictive performance Z-scores across tissue types (right).

The JTV method demonstrated superior accuracy and detail preservation. In brain parenchyma (gray and white matter), JTV achieved the highest predictive score with a mean Z-score of 1.19, significantly outperforming standard log-linear fits (-0.72) and adaptive optimized non-local means (AONLM, -0.41). Qualitatively, JTV was uniquely effective at preserving small structures like blood vessels, which are typically erased by traditional smoothing methods.

Estimated parameter maps (R1, MT, R2*) and their corresponding standard deviation (S.D.) maps across five repeat scans.

Furthermore, the method greatly reduced the variance of the estimated maps across repeated scans without introducing systematic bias. For example, in $R_2^*$ maps, the standard deviation was reduced by nearly half compared to non-regularized methods, leading to much cleaner quantitative parameter estimates.

Conclusion

By embedding spatial priors directly into the physical signal model, JTV-ESTATICS provides a robust framework for high-resolution qMRI. While the nonlinear MAP estimation is more computationally intensive than linear fitting, the gains in SNR and detail preservation are substantial. Future work may extend this generative framework to include super-resolution or patch-based priors to further enhance image quality.

Summary updated

Abstract

Quantitative magnetic resonance imaging (qMRI) derives tissue-specific parameters -- such as the apparent transverse relaxation rate R2*, the longitudinal relaxation rate R1 and the magnetisation transfer saturation -- that can be compared across sites and scanners and carry important information about the underlying microstructure. The multi-parameter mapping (MPM) protocol takes advantage of multi-echo acquisitions with variable flip angles to extract these parameters in a clinically acceptable scan time. In this context, ESTATICS performs a joint loglinear fit of multiple echo series to extract R2* and multiple extrapolated intercepts, thereby improving robustness to motion and decreasing the variance of the estimators. In this paper, we extend this model in two ways: (1) by introducing a joint total variation (JTV) prior on the intercepts and decay, and (2) by deriving a nonlinear maximum \emph{a posteriori} estimate. We evaluated the proposed algorithm by predicting left-out echoes in a rich single-subject dataset. In this validation, we outperformed other state-of-the-art methods and additionally showed that the proposed approach greatly reduces the variance of the estimated maps, without introducing bias.
BibTeX
@inproceedings{balbastre2020joint,
  author = {Yaël Balbastre and Mikael Brudfors and Michela Azzarito and Christian Lambert and Martina F. Callaghan and John Ashburner},
  title = {{Joint Total Variation ESTATICS for Robust Multi-Parameter Mapping}},
  booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year = {2020},
  url = {https://openreview.net/forum?id=nYYstWAF_yV},
}
Figure