Abstract
Cardiac Diffusion Tensor Imaging (DTI) provides critical insights into myocardial fiber architecture, yet its clinical utility is often limited by the requirement for prolonged breath-holding to mitigate motion artifacts. This paper details a robust method for acquiring DTI data during free-breathing using a temporal sampling strategy combined with a Principal Component Analysis and Temporal Maximum Intensity Projection (PCATMIP) algorithm. The approach successfully recovers signal lost to motion and provides more accurate DTI metrics in healthy volunteers.
Introduction
Understanding the orientation of myocardial fibers is essential for diagnosing and managing conditions like cardiomyopathy, as it reveals how the heart pumps and how disease alters its structure. Cardiac DTI maps the microscopic movement of water molecules to reveal this architecture. However, macroscopic motion from breathing and heartbeats causes significant signal attenuation and artifacts. While breath-holding is the standard solution, it is often impractical for patients with respiratory or cardiac distress. This research proposes a pipeline to achieve accurate DTI parameters without requiring patient cooperation in breath-holding.
Methodology
The proposed approach utilizes a two-stage strategy: specialized acquisition and advanced post-processing.
Temporal Sampling Strategy
Instead of a single snapshot, multiple Diffusion Weighted (DW) images are acquired across the diastolic phase. By incrementally increasing the "trigger delay" (TT) by 10ms for each successive frame, a temporal stack of images is generated. This ensures that the sequence captures the heart at various points in its cycle, including moments of relative quiescence.
PCATMIP Processing Pipeline
The raw temporal stack undergoes a three-step refinement process:
- Non-rigid Registration: Aligns images to correct for bulk respiratory-induced shifts.
- PCA Filtering: Enhances the signal-to-noise ratio (SNR) by isolating principal components of the signal.
- Temporal Maximum Intensity Projection (TMIP): Since motion typically reduces signal intensity in diffusion MRI, the TMIP algorithm selects the highest intensity pixels across the temporal stack. This effectively "cherry-picks" the moments where the heart was most still, recovering the true signal.
Results
The feasibility of the PCATMIP method was validated in healthy volunteers, demonstrating significant improvements over traditional single-trigger-delay (1TD) methods.
Signal Recovery and Metric Accuracy
The processing pipeline successfully recovered lost signal, leading to more reliable DTI parameters. Notably, the study observed a decrease in Fractional Anisotropy (FA) and Mean Diffusivity (MD) values post-processing, suggesting that raw data was previously inflated by motion artifacts.
- Fractional Anisotropy (FA): Reduced from $0.56 \pm 0.05$ (raw) to $0.43 \pm 0.05$ (processed).
- Mean Diffusivity (MD): Decreased from $1.37 \pm 0.16 \times 10^{-3} \text{ mm}^2/\text{s}$ to $0.75 \pm 0.12 \times 10^{-3} \text{ mm}^2/\text{s}$.
Furthermore, Helix Angle (HA) maps generated from the data showed the characteristic spiral pattern of healthy cardiac anatomy, confirming the structural integrity of the processed images.
Discussion and Conclusion
The ability to perform cardiac DTI during free-breathing represents a significant step toward clinical adoption. While the current scan time of approximately 20 minutes for 10 slices remains a hurdle for full clinical integration, the PCATMIP method provides a robust framework for motion-resistant imaging. Future work will focus on accelerating acquisition times and validating the method in patient populations with irregular heart rhythms, where motion compensation is even more critical. The authors conclude that this approach opens promising perspectives for mapping heart fiber health in patients who cannot tolerate breath-holding.
Summary updated