A Reconstruction-Agnostic Fan Phantom for Task-Relevant PET Spatial Resolution Assessment
Abstract
Purpose
Conventional approaches to PET spatial resolution assessment, such as the American College of Radiology (ACR) guideline and the NEMA NU 2-2018 protocol, offer limited insight into the effective spatial resolution achieved under clinical conditions. Modern PET imaging routinely relies on iterative reconstruction techniques, including OSEM and AI-enhanced methods and not filtered back projection. Additionally, fillable Derenzo-type phantoms provide only a discrete set of rod sizes. As a result, task-specific reconstruction optimization often proceeds without a practical tool to evaluate effective spatial resolution or to inform radiologists about the smallest reliably detectable feature.
Methods
We developed a reconstruction-agnostic method for assessing PET spatial resolution using a radioactivity-doped test pattern inspired by planar X-ray resolution phantoms. The phantom consists of a fan-shaped arrangement of line structures with continuously varying spatial frequency, enabling visual identification of the smallest resolvable feature at the point where adjacent lines can no longer be distinguished. The pattern is fabricated by cutting a fan geometry into paper, allowing preparation in under one minute. To achieve homogeneous activity distribution, the paper phantom is briefly soaked in a radioactive solution and sealed between two thin plastic sheets for imaging.
Results
The fan phantom provided a continuous range of line spacings, enabling rapid and intuitive identification of effective spatial resolution. Benchmarking against the NEMA NU-2 protocol produced comparable resolution estimates. The approach facilitated direct comparison across reconstruction methods, including filtered back projection, OSEM, and AI-based reconstructions. Unlike point-source metrics or Derenzo phantoms, the proposed method reflects clinically relevant, task-dependent image appearance.
Conclusion
This work presents a fast, practical, and reconstruction-agnostic approach for qualitative PET spatial resolution assessment. The simplicity and reproducibility of the phantom support on-demand use and potential integration into routine quality control workflows. Future work will address current limitations in contrast variability through multilayer phantom designs.