Abstract
Cardiac motion presents a significant challenge for in vivo diffusion tensor imaging (DTI) of the human heart, primarily due to substantial motion-induced signal loss. This study employs a multimodal simulation framework to quantitatively assess how cardiac motion biases diffusion tensor parameters. By integrating high-resolution fiber architecture from polarized light imaging (PLI) with realistic in vivo displacement fields from DENSE MRI, the research demonstrates that motion leads to a systematic overestimation of mean diffusivity (MD) and fractional anisotropy (FA). Furthermore, the study validates the efficacy of the PCATMIP (PCA filtering and temporal Maximum Intensity Projection) technique in recovering accurate structural information from motion-corrupted data.
Introduction
Diffusion Tensor Imaging (DTI) is a non-invasive technique capable of mapping the complex cardiomyocyte orientations within the living heart. However, the heart's continuous contraction and relaxation introduce artifacts that are often indistinguishable from actual water diffusion. This "motion-induced signal loss" complicates the measurement of key parameters like Fractional Anisotropy (FA) and Mean Diffusivity (MD), which are vital for assessing myocardial health. While the presence of these artifacts is well-known, quantifying their exact impact has been difficult due to the lack of a reliable "ground truth" for in vivo fiber architecture. This study addresses this gap by creating a sophisticated simulation environment that combines ex vivo structural data with in vivo motion dynamics.
Methods
The researchers developed a multimodal pipeline to isolate and quantify motion effects:
- Structural Ground Truth: High-resolution fiber orientation data was obtained from an ex vivo heart using Polarized Light Imaging (PLI).
- Realistic Motion Modeling: Displacement fields were captured from six healthy volunteers using DENSE (Displacement Encoding with Stimulated Echoes) MRI, providing a high-resolution map of actual tissue movement.
- Simulation and Artifact Injection: Monte-Carlo simulations generated Diffusion Weighted (DW) images based on the PLI data. These images were then "corrupted" using an empirical model that relates DENSE-derived displacement to signal intensity loss, alongside the addition of Rician noise.
- Correction Strategy: The PCATMIP technique—combining Principal Component Analysis (PCA) filtering with temporal Maximum Intensity Projection—was applied to the corrupted datasets to evaluate its ability to restore the original diffusion parameters.
Key Results
The simulation revealed that cardiac motion acts as a systematic bias rather than simple random noise, consistently inflating both diffusivity and anisotropy measurements.
- Mean Diffusivity (MD): The reference MD of $0.78 \times 10^{-3} \text{ mm}^2/\text{s}$ rose to $1.99 \pm 0.3 \times 10^{-3} \text{ mm}^2/\text{s}$ under motion. Application of PCATMIP successfully reduced this to $1.13 \pm 0.4 \times 10^{-3} \text{ mm}^2/\text{s}$, significantly closer to the ground truth.
- Fractional Anisotropy (FA): Motion increased the reference FA from $0.56$ to $0.61 \pm 0.05$. PCATMIP correction brought the value back to $0.59 \pm 0.02$.
- Fiber Architecture Recovery: Without correction, the characteristic rotation of fibers from the endocardium to the epicardium was lost. PCATMIP successfully recovered these azimuth angle variation patterns, measuring an elevation angle range ($41^\circ$ to $35^\circ$) that aligns with known physiological structures.
Discussion and Conclusion
This study provides a quantitative "motion tax" assessment for cardiac DTI. The results confirm that motion-induced signal loss leads to a significant overestimation of MD and FA, potentially masking or mimicking pathological changes in clinical settings. The multimodal approach—synergizing ex vivo PLI and in vivo DENSE—offers a robust platform for testing correction algorithms. While PCATMIP demonstrated a strong ability to recover structural patterns and improve parameter accuracy, the researchers noted that the corrected images maintain a higher noise floor than the reference data. Future work will focus on applying these validated correction techniques to clinical populations with heart disease, where accurate fiber mapping is critical for diagnosis and treatment planning.
Summary updated