Improving spatiotemporal resolution of USPIO-enhanced dynamic imaging of rat kidneys

Ying Sun, Dewen Yang, Qing Ye, Mangay Williams, José M. F. Moura, Fernando E. Boada, Zhi‐Pei Liang, Chien HoMagnetic 2003 · DOI

TL;DR

This paper introduces a novel imaging technique, Generalized Series (GS) imaging, to significantly improve the speed and resolution of dynamic MRI scans of rat kidneys enhanced with USPIO contrast agents. By cleverly reconstructing images from a reduced set of k-space data using pre- and post-contrast reference images, they achieve a substantial increase in temporal resolution without sacrificing image quality, paving the way for earlier detection of kidney transplant rejection.

What Problem Does This Paper Address?

Kidney transplantation is a vital treatment for end-stage renal disease, but graft failure remains a significant concern. Early detection of renal rejection is crucial for successful outcomes, and dynamic contrast-enhanced MRI is a promising tool for this purpose. However, conventional MRI techniques often struggle to achieve both high spatial and temporal resolution simultaneously, especially when imaging dynamic processes like contrast wash-in and wash-out in the kidneys. This limitation hinders the ability to capture rapid physiological changes and can lead to a loss of critical diagnostic information. Existing methods like keyhole imaging offer speed improvements but can suffer from artifacts due to data inconsistencies. This paper aims to bridge this gap by developing a method that enhances spatiotemporal resolution in USPIO-enhanced dynamic kidney MRI.

Methods

The core of the proposed method is a Generalized Series (GS) imaging scheme. Instead of reconstructing images directly from acquired k-space data using standard Fourier transforms, GS imaging expresses an unknown image as a linear combination of basis functions. These basis functions are constructed using a weighting function derived from reference images.

Specifically, the authors adapt a Two-Reference Iterative Generalized Series (TRIGR) reconstruction approach. The process involves:

  1. Acquiring two high-resolution reference images: One pre-contrast image (baseline) and one post-contrast image taken after the rapid dynamic wash-in phase.
  2. Acquiring reduced dynamic data sets: During the contrast agent's wash-in/wash-out period, only a subset of k-space data (e.g., 16 out of 64 phase encodings) is collected for each time frame.
  3. Reconstruction: The GS model uses the difference between the two reference images as a weighting function to guide the reconstruction of the dynamic images from the reduced data. This approach leverages the information from the high-resolution references to compensate for the missing high-frequency data in the dynamic scans.

The paper details a specific implementation where the contrast modulation function is derived from the low-resolution Fourier reconstruction of the dynamic data and the low-resolution version of the reference weighting function. A regularization term is introduced to ensure stable reconstruction.

Datasets

The paper utilizes experimental data acquired from Brown Norwegian rats undergoing USPIO-enhanced dynamic MRI of their kidneys.

  • "Gold" standard data: A sequence of 64 dynamic images with high spatial resolution (64x64 phase encodings) was acquired to serve as a benchmark.
  • Reduced data: The same k-space data was truncated to 16x16 phase encodings for each dynamic frame to simulate the reduced acquisition scenario.
  • Experimental data: Real-world experiments were conducted on rats using USPIO at a dose of 12.1 mg Fe/kg body weight, acquired on a 4.7-T Bruker AVANCE DRX MR instrument.

No new benchmarks were introduced; rather, the authors created reduced datasets from existing high-resolution data for simulation and used reduced acquisitions for experimental validation.

Key Insights and Ideas

  • Leveraging Reference Images is Key: The most significant insight is that by using both pre- and post-contrast high-resolution reference images, the GS reconstruction can effectively "fill in" the missing k-space information in the dynamic scans. This is a powerful data-sharing strategy.
  • Substantial Speedup with Minimal Quality Loss: The method demonstrates that it's possible to achieve a significant reduction in the number of acquired phase encodings (e.g., by a factor of 4 in 2D, potentially 16 in 3D) without a noticeable degradation in image quality compared to the "gold" standard.
  • Overcoming Keyhole Limitations: Unlike simpler keyhole methods, the GS approach is more robust to data inconsistencies between reference and dynamic data, leading to fewer artifacts and better image quality.
  • Preservation of Temporal Dynamics: Crucially, the TRIGR method accurately reproduces the temporal contrast uptake curves in different kidney tissues, indicating that the speedup does not introduce significant bias to the physiological measurements.

Key Results

  • Simulation Results: Images reconstructed using the proposed TRIGR method from truncated k-space data (16x16) showed significantly better spatial resolution and fewer blurring artifacts compared to conventional Fourier reconstruction from the same reduced data. The TRIGR images were comparable in quality to the "gold" standard (64x64).
  • Experimental Results: In a practical USPIO-enhanced dynamic imaging study of rat kidneys, the TRIGR method successfully reconstructed dynamic images with 16 phase encodings, effectively removing the severe blurring artifacts seen in conventionally reconstructed images.
  • Temporal Response: Analysis of temporal intensity curves showed that TRIGR accurately captured the contrast uptake in cortical and muscle tissues, matching the results from conventional Fourier reconstruction, despite the drastic reduction in acquired data.

Limitations and Future Work

  • Limitations:
    • The paper notes that the regularization constant ($\lambda$) in the reconstruction needs to be carefully chosen to balance stability and accuracy.
  • While the method significantly improves resolution, it's still a trade-off; the "gold" standard still represents the ultimate resolution.
  • The current implementation is demonstrated in 2D, with potential for 3D applications.
  • Future Work:
    • The authors suggest that this method is expected to make 3D studies possible using USPIO-enhanced dynamic imaging of rat kidneys.
  • The technique is anticipated to be valuable for early detection of renal rejection after kidney transplantation due to its improved speed and resolution.
  • Further exploration of the optimal choice of basis functions and regularization parameters could be beneficial.

Summary updated

BibTeX
@inproceedings{sun2003improving,
  author = {Ying Sun and Dewen Yang and Qing Ye and Mangay Williams and José M. F. Moura and Fernando E. Boada and Zhi‐Pei Liang and Chien Ho},
  title = {{Improving spatiotemporal resolution of USPIO-enhanced dynamic imaging of rat kidneys}},
  booktitle = {Magnetic},
  year = {2003},
  doi = {10.1016/s0730-725x(03)00089-4},
}
Figure