Brightness-Invariant Tracking Estimation in Tagged MRI

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Magnetic resonance (MR) tagging is a noninvasive technique to track tissue motion in vivo.
• Conventional optical flow and Fourier methods for tagged MR image analysis are susceptible to errors caused by changing tag and tissue brightness over time.
• Tag fading and spectral spreading due to motion further complicate accurate motion tracking.

Purpose of the Study:

• To introduce a novel technique, brightness-invariant tracking estimation (BRITE), for accurate motion tracking in tagged MRI.
• To address the limitations of existing methods in handling brightness variations and tag fading.
• To disentangle anatomy from tag patterns and simultaneously estimate Lagrangian motion.

Main Methods:

• BRITE utilizes denoising diffusion probabilistic models to represent anatomy and physics-informed neural networks for biologically plausible motion estimation.
• The technique disentangles anatomical information from the tag pattern within tagged MR image sequences.
• Experiments were conducted on a gel phantom with varying tag periods and flip angles to assess performance under different brightness conditions.

Main Results:

• BRITE demonstrated superior accuracy in motion and strain estimation compared to state-of-the-art methods.
• The proposed method exhibits significant resistance to tag fading, a common artifact in tagged MRI.
• Validation using a gel phantom confirmed the effectiveness of BRITE in diverse imaging scenarios.

Conclusions:

• BRITE offers a robust solution for accurate motion and strain quantification in tagged MRI, overcoming challenges posed by brightness variations and tag fading.
• The integration of deep learning models (diffusion models and physics-informed neural networks) enhances the reliability and accuracy of the technique.
• BRITE represents a significant advancement for in vivo tissue motion analysis using tagged MR imaging.