Brightness-Invariant Tracking Estimation in Tagged MRI

TL;DR

BRITE tackles brightness-invariant tracking in tagged MRI by separating the anatomical signal from tag patterns and by explicitly modeling tag fading over time. Its two-stage framework uses a pretrained denoising diffusion probabilistic model as an anatomical prior and a physics-informed neural network to estimate Lagrangian motion, while a tag fading module adapts the tag pattern parameters frame-by-frame. Evaluation on SPAMM-tagged silicone gel phantom data across multiple tag periods and flip angles shows BRITE yields more accurate motion and strain estimates than state-of-the-art methods, particularly under brightness changes. BRITE reduces sensitivity to tag fading and spectral overlap and avoids error accumulation by sequential initialization, providing a practical approach for reliable motion tracking in tagged MRI.

Abstract

Magnetic resonance (MR) tagging is an imaging technique for noninvasively tracking tissue motion in vivo by creating a visible pattern of magnetization saturation (tags) that deforms with the tissue. Due to longitudinal relaxation and progression to steady-state, the tags and tissue brightnesses change over time, which makes tracking with optical flow methods error-prone. Although Fourier methods can alleviate these problems, they are also sensitive to brightness changes as well as spectral spreading due to motion. To address these problems, we introduce the brightness-invariant tracking estimation (BRITE) technique for tagged MRI. BRITE disentangles the anatomy from the tag pattern in the observed tagged image sequence and simultaneously estimates the Lagrangian motion. The inherent ill-posedness of this problem is addressed by leveraging the expressive power of denoising diffusion probabilistic models to represent the probabilistic distribution of the underlying anatomy and the flexibility of physics-informed neural networks to estimate biologically-plausible motion. A set of tagged MR images of a gel phantom was acquired with various tag periods and imaging flip angles to demonstrate the impact of brightness variations and to validate our method. The results show that BRITE achieves more accurate motion and strain estimates as compared to other state of the art methods, while also being resistant to tag fading.

Figures (7)

• Figure 1: Top row: Tagged MRI. Middle Row: Their corresponding Fourier spectrum. Bottom Row: The midline profiles extracted from the Fourier spectra. The column names "Xmm@Ys" denote a tag period of X mm acquired at Y s following tagging.
• Figure 2: Tagged MRI of a static phantom. Harmonic phases are distorted over time due to tag fading and spectral overlap, resulting in erroneous non-zero motion estimation.
• Figure 3: Overview of the Proposed BRITE Framework. The pipeline consists of two stages: disentangling and tracking. At $t = 0$, the "Tag Pattern Module" and a pretrained DDPM model are used to separate the anatomical image $\tilde{a}_{0}$ from the tag patterns $\tilde{p}^{h,v}_{0}$. Subsequently, for each time $t>0$, the PINN estimates the Lagrangian motion while the "Tag Fading Module" captures tag fading.
• Figure 4: (a) Input tagged MR sequence with horizontal and vertical tags. (b) Synthetic oval-shaped images used for training the DDPM. (c) Disentanglement process: shown are the optimization trajectories of $\mu, \varphi^h, \varphi^v, A_0, B_0$, the disentanglement loss, estimated anatomy $\tilde{a}_0$, and reconstructed tagged images $\tilde{g}_0$ at 50 and 600 iterations. (d) Tag fading module optimization. Every 5th frame (total of 20 frames over 1.1s) is used for optimization and display clarity. (e) Estimated Lagrangian motion fields.
• Figure 5: Non-rigid Deformation Evaluation. Results are shown for EPE (top two rows) and eMPS (bottom two rows) at different time points. Each column represents a different tag period, and each row corresponds to a flip angle.