List-mode-based reconstruction for respiratory motion correction in PET using non-rigid body transformations

F. Lamare, María J. Ledesma‐Carbayo, Thierry Cresson, George Kontaxakis, A. Santos, Catherine Cheze Le Rest, Andrew J. Reader, Dimitris VisvikisPhysics 2007 · DOI

TL;DR

This paper introduces a novel method for correcting respiratory motion artifacts in PET imaging by integrating non-rigid (elastic) transformations directly into the list-mode reconstruction process. This approach significantly improves image quality, especially in the lungs, by utilizing all available data and achieving more uniform correction compared to previous methods.

What Problem Does This Paper Address?

Respiratory motion is a major challenge in PET imaging, particularly for thoracic and abdominal organs. This motion blurs images, reduces lesion contrast, and can lead to inaccurate quantitative measurements, impacting crucial applications like radiotherapy planning and treatment monitoring. Current solutions often involve "gating" – acquiring data in separate frames synchronized with breathing. However, this approach results in frames with low statistics and poor signal-to-noise ratios. The paper aims to overcome these limitations by developing a method that corrects for respiratory motion during the reconstruction of the entire PET dataset, rather than relying on pre-reconstruction processing or summing of low-statistic gated frames.

Methods

The core of this work lies in integrating elastic transformations into a list-mode Expectation-Maximization (EM) reconstruction algorithm, specifically the One-Pass List Mode EM (OPL-EM).

  • Non-Rigid Registration: The authors first used a non-rigid registration algorithm based on B-spline basis functions to derive transformation fields from dynamic CT images of a realistic anthropomorphic phantom (NCAT). These fields describe how different parts of the anatomy move throughout the respiratory cycle.
  • List-Mode Reconstruction with Elastic Transformations: The key innovation is incorporating these derived elastic transformations directly into the OPL-EM algorithm's system matrix. This means that as the algorithm iterates to reconstruct the PET image, it accounts for the motion of each voxel at each time point.
  • Two Implementations: Two distinct methods for incorporating these elastic transformations were explored:
  • Elastic Method 1 (Interpolation-based): This method uses trilinear interpolation to calculate the system matrix coefficients, allowing the use of the efficient Siddon ray tracing algorithm.
  • Elastic Method 2 (Direct Calculation): This method directly calculates the coefficients based on the exact transformed voxel positions, avoiding interpolation but leading to longer reconstruction times.
  • Comparison: The performance of these elastic methods was compared against:
  • An uncorrected "average" image (sum of all frames without correction).
  • An image corrected using an affine (rigid) transformation applied before reconstruction.
  • An image reconstructed from gated frames that were individually corrected and then summed.

Datasets

The study utilized a digital NURBS-based 4D cardiac-torso phantom (NCAT) for simulation. This phantom was designed to realistically model respiratory motion, including non-rigid deformations in the lungs and diaphragm. Various lesion sizes (7, 11, 15, and 21 mm) were included at different locations within the lungs and liver. Simulated PET data was generated by combining NCAT emission images with Monte Carlo simulations to account for physical processes and attenuation. No real patient data was used for training or evaluation in this specific study.

Key Insights and Ideas

  • Direct Integration is Superior: Incorporating motion correction during reconstruction (as done with the elastic methods) yields significantly better results than applying transformations to already reconstructed images and then summing them. This is because it preserves image statistics and avoids the loss of information that occurs when summing low-statistic gated frames.
  • Non-Rigid Motion is Crucial for Thorax: Simple affine transformations are insufficient for accurately correcting respiratory motion in the thorax, especially for organs under the diaphragm. Elastic transformations are necessary to capture the complex, non-uniform movements.
  • Interpolation Trade-offs: While the interpolation-based elastic method (Method 1) offers faster reconstruction, it introduces artefacts and reduces quantitative improvements, particularly in areas with large deformations like the diaphragm. The direct calculation method (Method 2) provides superior results but at the cost of significantly longer processing times.
  • Uniformity of Correction: The elastic methods, especially the direct calculation approach, provide a more uniform improvement in lesion contrast and position across different lung regions and lesion sizes compared to affine methods.

Key Results

  • Contrast Improvement: Elastic correction (specifically Method 2) achieved contrast improvements of 85-99% for lung lesions, significantly outperforming affine correction (which showed improvements of up to 30% for lung lesions and less for liver lesions).
  • Position and FWHM Improvement: Elastic methods demonstrated substantial improvements in lesion position and Full Width at Half Maximum (FWHM), with improvements ranging from 65% to 99% along the Y-axis and significant gains along the Z-axis.
  • Artefacts with Interpolation: The interpolation-based elastic method (Method 1) showed some artefacts, particularly in the diaphragm region, leading to a loss of voxel intensity and reduced quantitative improvements compared to the direct calculation method.
  • Reconstruction Time: The direct calculation elastic method (Method 2) was over 10 times slower than the interpolation-based method.

Limitations and Future Work

  • Reconstruction Time: The primary limitation of the superior direct calculation elastic method is its significantly increased reconstruction time.
  • Interpolation Artefacts: The interpolation-based method introduces artefacts, especially in areas of large deformation, limiting its quantitative performance.
  • Future Work: The authors suggest that further research could focus on optimizing the reconstruction speed of the direct calculation method or developing more robust interpolation techniques. Investigating the application of these methods to real patient data is also a logical next step. The paper also notes that neither method explicitly accounts for deformation of voxel shape, which could be a consideration for future improvements.

Summary updated

Abstract

Respiratory motion in emission tomography leads to reduced image quality. Developed correction methodology has been concentrating on the use of respiratory synchronized acquisitions leading to gated frames. Such frames, however, are of low signal-to-noise ratio as a result of containing reduced statistics. In this work, we describe the implementation of an elastic transformation within a list-mode-based reconstruction for the correction of respiratory motion over the thorax, allowing the use of all data available throughout a respiratory motion average acquisition. The developed algorithm was evaluated using datasets of the NCAT phantom generated at different points throughout the respiratory cycle. List-mode-data-based PET-simulated frames were subsequently produced by combining the NCAT datasets with Monte Carlo simulation. A non-rigid registration algorithm based on B-spline basis functions was employed to derive transformation parameters accounting for the respiratory motion using the NCAT dynamic CT images. The displacement matrices derived were subsequently applied during the image reconstruction of the original emission list mode data. Two different implementations for the incorporation of the elastic transformations within the one-pass list mode EM (OPL-EM) algorithm were developed and evaluated. The corrected images were compared with those produced using an affine transformation of list mode data prior to reconstruction, as well as with uncorrected respiratory motion average images. Results demonstrate that although both correction techniques considered lead to significant improvements in accounting for respiratory motion artefacts in the lung fields, the elastic-transformation-based correction leads to a more uniform improvement across the lungs for different lesion sizes and locations.
BibTeX
@inproceedings{lamare2007list,
  author = {F. Lamare and María J. Ledesma‐Carbayo and Thierry Cresson and George Kontaxakis and A. Santos and Catherine Cheze Le Rest and Andrew J. Reader and Dimitris Visvikis},
  title = {{List-mode-based reconstruction for respiratory motion correction in PET using non-rigid body transformations}},
  booktitle = {Physics},
  year = {2007},
  doi = {10.1088/0031-9155/52/17/006},
}
Figure