Design Optimisation of a Flat-Panel, Limited-Angle TOF-PET Scanner: A Simulation Study

M. Orehar, R. Dolenec, Georges El Fakhri, S. Korpar, P. Križan, Gašper Razdevšek, Thibault Marin, Dejan Žontar, R. PestotnikDiagnostics 2024 · DOI

TL;DR

This study explores the potential of flat-panel Time-of-Flight Positron Emission Tomography (TOF-PET) scanners, demonstrating through simulations that they can achieve image quality comparable to conventional ring scanners with significantly less scintillator material. By leveraging improved coincidence time resolution (CTR), these novel designs offer a more cost-effective and compact alternative for PET imaging.

What Problem Does This Paper Address?

Traditional PET scanners use a ring geometry, which requires full angular coverage and thus a large amount of scintillator material and associated costs. However, advancements in TOF technology, specifically achieving a better coincidence time resolution (CTR), relax the need for complete angular coverage. This opens up possibilities for alternative scanner designs, such as flat-panel configurations. The challenge lies in optimizing these flat-panel designs to match the image quality of established clinical scanners while capitalizing on their inherent advantages of reduced cost, compactness, and modularity. This paper investigates how various design parameters of a four-panel flat-panel TOF-PET scanner, including crystal dimensions, readout strategies, and CTR, influence performance metrics like spatial resolution, image quality, and sensitivity.

Methods

The researchers employed Monte Carlo simulations using the GATE software (v8.2) to model different flat-panel PET scanner configurations. Image reconstruction was performed using the CASTOR software (v3.1.1). The study simulated various scanner designs, varying crystal cross-sections (from 1.5x1.5 mm² to 3x3 mm²) and lengths (5 mm to 10 mm), and tested different coincidence time resolutions (CTRs) ranging from 75 ps to 200 ps.

Key methodological aspects include:

  • Scanner Geometry: A four-panel flat-panel configuration was simulated, with panels spaced 40 cm apart.
  • Point Source Spatial Resolution Evaluation: A novel approach was used to assess spatial resolution across the entire field of view (FOV) by simulating 125 point sources in a grid and fitting the reconstructed images with a Gaussian function, moving beyond the limitations of the standard NEMA method for non-cylindrical geometries.
  • Image Quality Assessment: The NEMA phantom was used to evaluate image quality, measuring contrast recovery and background variability.
  • Sensitivity and NECRTOF: Standard metrics were used to quantify scanner sensitivity and noise-equivalent count rate (NECRTOF), with a focus on how these are affected by design choices.
  • Readout Strategies: Both single-crystal and submodule (5x5 crystals) readout strategies were investigated to understand their impact on performance.
  • Reference Scanner: A Siemens Biograph Vision scanner was used as a benchmark for comparison.

Datasets

The study did not use pre-existing datasets for training. Instead, it relied on simulated data generated using GATE. The evaluation was performed using:

  • Point Sources: 125 spherical point sources arranged in a grid within the FOV to assess spatial resolution.
  • NEMA Phantom: A standard phantom used for evaluating image quality metrics like contrast recovery and background variability.
  • Brain Phantom: A high-resolution digital brain phantom (BigBrain atlas) was used for qualitative assessment of image quality.

Key Insights and Ideas

  • CTR is King: The study strongly emphasizes that achieving a sufficiently low CTR (around 75 ps) is crucial for flat-panel scanners. This improved timing resolution significantly compensates for the reduced angular coverage inherent in flat-panel designs, enabling comparable image quality to conventional scanners.
  • Cost-Effectiveness: Flat-panel designs, even with optimized parameters, use substantially less scintillator material (as little as 28.9% compared to the reference scanner), directly translating to lower manufacturing costs.
  • Modularity and Compactness: The flat-panel geometry inherently offers greater modularity and a more compact form factor, potentially leading to more mobile and versatile PET systems.
  • Spatial Resolution Trade-offs: While reducing crystal cross-section can improve spatial resolution, it also leads to decreased sensitivity. Crystal length has a more direct impact on spatial resolution, with shorter crystals generally performing better, highlighting the potential importance of Depth of Interaction (DOI) capabilities in future designs.
  • Readout Strategy Matters: Submodule readout generally offers better sensitivity and image quality compared to single-crystal readout, although single-crystal readout can provide higher spatial resolution. The ability to utilize both strategies could offer a powerful combination of benefits.
  • Beyond NEMA: The paper highlights the need for adapted methods to assess spatial resolution in non-cylindrical geometries like flat-panel scanners, proposing a Gaussian fitting approach for point sources.

Key Results

  • Image Quality: The best-performing flat-panel configuration (CS3_L10_CTR75) achieved image quality comparable to the reference Siemens Biograph Vision scanner, even with significantly less scintillator material.
  • Spatial Resolution: The optimized flat-panel scanner demonstrated better spatial resolution across most of the FOV compared to the reference scanner, with measured resolutions ranging from 1.6 mm to 2.7 mm in the best case.
  • Sensitivity: While the reference scanner generally showed higher sensitivity, the submodule readout for the flat-panel scanner significantly improved its sensitivity, approaching that of the reference.
  • Cost Reduction: The optimized flat-panel design used only about 28.9% of the scintillator material of the reference scanner, indicating substantial cost savings.
  • CTR Impact: Simulations showed that improving CTR from 100 ps to 75 ps led to noticeable improvements in image quality metrics.

Limitations and Future Work

  • Simulation-Based: The study is entirely based on simulations. While rigorous, real-world implementation may introduce unforeseen challenges.
  • Optimized CTR Assumption: The study assumes a system-level CTR of 75 ps, which is projected to be achievable in the near future. The actual realization of this CTR in clinical systems remains a key factor.
  • Limited Axial FOV: While flat-panel designs offer modularity for extending axial coverage, this study focused on a four-panel configuration with a specific axial FOV. Future work will explore larger axial coverage configurations.
  • DOI Not Explicitly Modeled: While the importance of DOI for spatial resolution is mentioned, it was not explicitly modeled in this simulation study.
  • Computational Cost: The study acknowledges the significant computational cost of simulations, especially for detailed spatial resolution analysis with background activity. This necessitates efficient computing resources and potentially alternative analysis methods for future research.

The authors suggest future research will focus on exploring larger axial coverage configurations and further investigating the benefits of Depth of Interaction (DOI) capabilities. They also propose the development of scanners that can support both crystal and submodule readout strategies to leverage the advantages of each.

Summary updated

Abstract

In time-of-flight positron emission tomography (TOF-PET), a coincidence time resolution (CTR) below 100 ps reduces the angular coverage requirements and, thus, the geometric constraints of the scanner design. Among other possibilities, this opens the possibility of using flat-panel PET detectors. Such a design would be more cost-accessible and compact and allow for a higher degree of modularity than a conventional ring scanner. However, achieving adequate CTR is a considerable challenge and requires improvements at every level of detection. Based on recent results in the ongoing development of optimised TOF-PET photodetectors and electronics, we expect that within a few years, a CTR of about 75 ps will be be achievable at the system level. In this work, flat-panel scanners with four panels and various design parameters were simulated, assessed and compared to a reference scanner based on the Siemens Biograph Vision using NEMA NU 2-2018 metrics. Point sources were also simulated, and a method for evaluating spatial resolution that is more appropriate for flat-panel geometry is presented. We also studied the effects of crystal readout strategies, comparing single-crystal and module readout levels. The results demonstrate that with a CTR below 100 ps, a flat-panel scanner can achieve image quality comparable to that of a reference clinical scanner, with considerable savings in scintillator material.
BibTeX
@inproceedings{orehar2024design,
  author = {M. Orehar and R. Dolenec and Georges El Fakhri and S. Korpar and P. Križan and Gašper Razdevšek and Thibault Marin and Dejan Žontar and R. Pestotnik},
  title = {{Design Optimisation of a Flat-Panel, Limited-Angle TOF-PET Scanner: A Simulation Study}},
  booktitle = {Diagnostics},
  year = {2024},
  doi = {10.3390/diagnostics14171976},
}
Figure