Accelerated Motion-Aware MR Imaging via Motion Prediction from K-Space Center

Christoph Jud, Damien Nguyen, Alina Giger, Robin Sandkühler, Miriam Krieger, Tony Lomax, Rares Salomir, Oliver Bieri, Philippe C. CattinarXiv 2019 · arXiv

Abstract

Magnetic Resonance Imaging (MRI) of the chest and abdomen is frequently compromised by respiratory motion, leading to artifacts that necessitate costly repeat scans. This paper introduces an accelerated 4D MRI technique that predicts complex, non-rigid motion using a fraction of the typical data requirements. By integrating a brief training phase with a cubic regression model, the proposed approach reduces acquisition times by half and reconstruction times from two hours to approximately three minutes, while maintaining sub-pixel motion prediction accuracy and enhancing image sharpness.

Introduction

MRI is a powerful diagnostic tool but is inherently slow, making it susceptible to patient motion. In thoracic and abdominal imaging, breathing causes blurring and ghosting artifacts. Existing motion-aware techniques often rely on continuous oversampling of the k-space center to track motion, but these methods typically require long acquisition times (over 11 minutes) and extensive post-processing, limiting their clinical utility.

Methodology

The authors propose a two-phase accelerated motion-aware approach:

  1. Training Phase: For approximately 76 seconds, the scanner captures a large portion of the k-space center. This data is used for 3D image registration to model the patient's unique non-rigid breathing patterns.
  2. Inference Phase: During the subsequent 4.5 minutes, the scanner captures only a minimal subset of the k-space center (reducing the sampling radius from 10 to 2 points).
  3. Motion Prediction: Signals from the minimal k-space subset are mapped to the full motion fields learned during training using Cubic Regression and Principal Component Analysis (PCA).
  4. Temporal Alignment: To correct for the time delay between center (motion) and periphery (image) data acquisition, a quadratic interpolation shifts the motion field by 0.5 time steps, ensuring precise alignment.
The proposed workflow splits the scan into a training phase to learn motion patterns and an inference phase for high-speed data acquisition.

Experimental Results

The method was validated on 12 healthy volunteers using 1.5T and 3T MRI scanners. Key performance metrics include:

  • Acquisition Speed: Total scan time was reduced from 11.1 minutes to 5.8 minutes.
  • Reconstruction Efficiency: Processing time dropped from 2 hours to roughly 3 minutes, enabling near real-time clinical feedback.
  • Prediction Accuracy: The average motion prediction error remained below 2 mm across most subjects, representing sub-pixel accuracy.
  • Image Quality: Statistical analysis indicated a significant increase in Total Variation, reflecting improved image sharpness due to the temporal shift correction.
Prediction error across 12 volunteers, showing mean and 95th percentile values. Volunteer 9 represents an outlier due to breathing drift.

Discussion and Limitations

The primary challenge identified is "breathing drift," where a patient's respiratory pattern changes significantly after the training phase. This phenomenon was observed in Volunteer 9, leading to higher prediction errors. Future iterations could involve training the model on a wider range of motion (e.g., guided deep breaths) to improve generalization.

Conclusion

By leveraging the redundancy in k-space center data and employing predictive modeling, this research demonstrates a path toward high-speed, motion-robust 4D MRI. The dramatic reduction in both acquisition and reconstruction times brings advanced motion-aware imaging closer to routine clinical practice.

Summary updated

Abstract

Motion has been a challenge for magnetic resonance (MR) imaging ever since the MR has been invented. Especially in volumetric imaging of thoracic and abdominal organs, motion-awareness is essential for reducing motion artifacts in the final image. A recently proposed MR imaging approach copes with motion by observing the motion patterns during the acquisition. Repetitive scanning of the k-space center region enables the extraction of the patient motion while acquiring the remaining part of the k-space. Due to highly redundant measurements of the center, the required scanning time of over 11 min and the reconstruction time of 2 h exceed clinical applicability though. We propose an accelerated motion-aware MR imaging method where the motion is inferred from small-sized k-space center patches and an initial training phase during which the characteristic movements are modeled. Thereby, acquisition times are reduced by a factor of almost 2 and reconstruction times by two orders of magnitude. Moreover, we improve the existing motion-aware approach with a systematic temporal shift correction to achieve a sharper image reconstruction. We tested our method on 12 volunteers and scanned their lungs and abdomen under free breathing. We achieved equivalent to higher reconstruction quality using the motion-prediction compared to the slower existing approach.
BibTeX
@misc{jud2019accelerated,
  author = {Christoph Jud and Damien Nguyen and Alina Giger and Robin Sandkühler and Miriam Krieger and Tony Lomax and Rares Salomir and Oliver Bieri and Philippe C. Cattin},
  title = {{Accelerated Motion-Aware MR Imaging via Motion Prediction from K-Space Center}},
  year = {2019},
  eprint = {1908.09560},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1908.09560},
}
Figure