Abstract
Tagged MRI motion correction in simultaneous PET/MRI significantly improves lesion detection compared with respiratory gating and no motion correction while reducing radiation dose. In vivo primate and rabbit studies confirmed the improvement in PET image quality and provide the rationale for evaluation in simultaneous whole-body PET/MRI clinical studies.
Introduction
Positron Emission Tomography (PET) is a cornerstone of oncological imaging, yet it suffers from a fundamental "shutter speed" limitation. Because scans typically last several minutes, physiological motion—primarily respiratory and cardiac—blurs the resulting images. This motion can degrade spatial resolution from a theoretical 4–7 mm to approximately 11 mm in thoracic and abdominal regions, potentially masking small lesions and complicating quantification.
Traditional solutions involve physiologic gating, which "freezes" motion by only utilizing data from a specific phase of the respiratory cycle. However, this approach discards 80-90% of the acquired signal, leading to high noise levels and poor signal-to-noise ratios (SNR). The authors sought a method to use 100% of the detected photons without the associated blur.
Methodology: Simultaneous Nonrigid Correction
To overcome these limitations, researchers developed a method for MRI-based nonrigid motion correction using a simultaneous PET/MRI system. This workflow integrates three core components:
- Tagged MRI (SPAMM): A Spatial Modulation of Magnetization sequence "paints" a temporary magnetic grid onto the tissue. As organs move and deform, the grid tracks these complex, nonrigid changes in real-time.
- Motion Estimation: 2D B-spline nonrigid image registration calculates precise "motion fields" for every voxel throughout the respiratory cycle.
- Motion-Corrected Reconstruction: These motion fields are incorporated directly into the List-mode OSEM PET reconstruction algorithm. Every detected photon is mathematically "warped" back to a reference frame, ensuring that 100% of the acquired data contributes to a sharp, motion-free image.
Quantitative Performance and Lesion Detection
The efficacy of this approach was validated using a Channelized Hotelling Observer (CHO) model—a mathematical model of human signal detection—to assess lesion detection sensitivity. In phantom studies, MRI-based correction demonstrated a dramatic improvement in detection SNR, ranging from 166% to 276% compared to standard gating.
For significant motions (e.g., ~8.8 mm), the method improved lesion contrast by up to 58% over uncorrected images. Crucially, the motion-corrected images achieved an SNR equivalent to gated images that would require 5 to 8 times longer acquisition periods, representing a massive leap in imaging efficiency.
Visual Clarity and Resolution
The method successfully breaks the traditional trade-off between noise and resolution. By re-centering every detected event, the reconstruction achieves the sharpness of a gated scan with the low noise profile of a full-duration acquisition.
As shown in the intensity profiles, the MRI motion-corrected reconstruction (green line) significantly recovers peak intensity compared to the uncorrected scan (orange line). It matches the sharpness of a 70-minute gated scan (purple line) while using only 5 minutes of data, far outperforming the noisy 5-minute gated alternative (black line).
Conclusion
By leveraging the simultaneity of PET/MRI, this nonrigid motion correction technique provides a path toward high-resolution, low-noise PET imaging without the dose or time penalties of traditional gating. While computational demands remain high—taking approximately one hour per iteration in primate studies—future implementation on GPU-based architectures promises to bring this high-sensitivity detection into routine clinical practice.
Summary updated