Accurate 3D Positron Range Correction Method for Heterogeneous Material Densities in PET

Chong Li, J. Scheins, Lutz Tellmann, Ahlam Issa, Long Wei, N. Jon Shah, Christoph LerchemedRxiv 2022 · DOI

Abstract

The positron range is a fundamental, detector-independent physical limitation to spatial resolution in positron emission tomography (PET), causing significant blurring in reconstructed images. A major challenge for correction methods is providing accurate range kernels that incorporate inhomogeneous stopping power, particularly at tissue boundaries. This work proposes a novel approach to generate accurate three-dimensional (3-D) blurring kernels in both homogeneous and heterogeneous media. By tracking positron energy deposition along straight paths based on material stopping power—derived directly from 511 keV gamma attenuation maps—the method accounts for energy deposition history within range kernels. Validated against Geant4 simulations and physical measurements in polyurethane phantoms, the proposed method shows high agreement (annihilation probability differences <20%) and correctly models tissue boundaries. This approach significantly improves spatial resolution for non-standard isotopes even in cases of significant tissue inhomogeneities.

Introduction: The Challenge of Positron Range

In Positron Emission Tomography (PET), the goal is to pinpoint the exact location of a radiotracer. However, the physics of positron decay introduces a fundamental "blurring" effect known as the positron range. When a radioactive atom decays, it emits a positron that travels a short distance before colliding with an electron and annihilating. The scanner detects this annihilation point, which may be several millimeters (or even centimeters for high-energy isotopes like $^{68}$Ga or $^{120}$I) away from the original decay site.

Correcting for this range is particularly difficult in heterogeneous environments. A positron’s path is dictated by the specific materials it traverses—such as moving from soft tissue into dense bone or airy lung. Existing correction methods often rely on slow Monte Carlo simulations or oversimplified models that ignore the specific path history of the positron.

The Quasi-Continuous Energy Loss (CEL) Method

The researchers developed the Quasi-continuous energy loss (CEL) method to provide a fast, accurate solution for 3D resolution recovery. The core innovation lies in linking the positron stopping power directly to the linear attenuation coefficient ($\mu_L$) of 511 keV gamma photons. Because every PET scan already generates an "attenuation map" to correct for photon absorption, the CEL method can leverage this existing data to calculate positron range on-the-fly without requiring additional CT segmentation or manual tissue labeling.

Geometric implementation of the CEL method and lookup table (LUT) generation.

To ensure clinical feasibility, the method utilizes precomputed Look-Up Tables (LUTs) for object-independent geometric data, such as path IDs and voxel intersections. This allows the final range kernel to be generated in under 10 minutes, a significant improvement over traditional simulation times.

Experimental Validation

The CEL method was validated through a combination of gold-standard Geant4 Monte Carlo simulations and physical experiments. The researchers utilized custom-made phantoms constructed from four different densities of low-density polyurethane (ranging from 0.01 to 0.089 g/cm³). These low densities effectively "stretch" the positron range of $^{18}$F, allowing for precise measurement of depth profiles using a high-resolution 3T MR-BrainPET scanner.

Physical phantoms and experimental setup used to validate the CEL method.

Key Results and Insights

The study yielded several critical findings:

  • High Fidelity: Annihilation probabilities calculated via the CEL method showed less than a 20% difference compared to computationally expensive Geant4 simulations.
  • Boundary Accuracy: The model successfully handled sharp transitions at tissue boundaries (e.g., water-to-bone and water-to-lung), which typically cause artifacts in simpler models.
  • Path History Significance: By accounting for the specific order of tissues a positron traverses, the CEL method provides a more physically accurate representation of range than "average density" models.
  • Numerical Stability: Residual artifacts from numerical computations were kept remarkably low, at approximately 1%.

Limitations and Future Directions

While highly effective, the current CEL method employs a straight-path approximation. In reality, positrons follow a "zig-zag" path due to multiple scattering events. Although the authors integrated a correction factor for this behavior, it remains an approximation of the true 3D annihilation cloud.

Furthermore, the current model does not yet account for the strong magnetic fields found in hybrid PET/MRI scanners, which can spiral positrons and alter their range. Future iterations of the CEL method aim to become "magnetic field-aware" to support the next generation of high-resolution hybrid imaging.

Summary updated

Abstract

Abstract Objective The positron range is a fundamental, detector-independent physical limitation to special resolution in positron emission tomography (PET) as it causes a significant blurring of the reconstructed PET images. A major challenge for positron range correction methods is to provide accurate range kernels that inherently incorporate the generally inhomogeneous stopping power, especially at tissue boundaries. In this work, we propose a novel approach to generate accurate three-dimensional (3-D) blurring kernels both in homogenous and heterogeneous media to improve PET spatial resolution. Approach In the proposed approach, positron energy deposition was approximately tracked along straight paths, depending on the positron stopping power of the underlying material. The positron stopping power was derived from the attenuation coefficient of 511keV gamma photons according to the available PET attenuation maps. Thus, the history of energy deposition is taken into account within the range of kernels. Special emphasis was placed on facilitating the very fast computation of the positron annihilation probability in each voxel. Results Positron path distributions of 18 F in low-density polyurethane were in high agreement with Geant4 simulation at an annihilation probability larger than 10 −2 ∼10 −3 of the maximum annihilation probability. The Geant4 simulation was further validated with measured 18 F depth profiles in these polyurethane phantoms. The tissue boundary of water with cortical bone and lung was correctly modeled. Residual artifacts from the numerical computations were in the range of 1%. The calculated annihilation probability in voxels shows an overall difference of less than 20% compared to the Geant4 simulation. Significance The proposed method significantly improves spatial resolution for non-standard isotopes by providing accurate range kernels, even in the case of significant tissue inhomogeneities.
BibTeX
@misc{li2022accurate,
  author = {Chong Li and J. Scheins and Lutz Tellmann and Ahlam Issa and Long Wei and N. Jon Shah and Christoph Lerche},
  title = {{Accurate 3D Positron Range Correction Method for Heterogeneous Material Densities in PET}},
  year = {2022},
  doi = {10.1101/2022.08.16.22278715},
  howpublished = {medRxiv},
}

Loading related content...

Figure