PhD Scientific Days 2026

Poster Session 1.I - Theoretical and Translational Medicine

A Simple, Dynamic Geometric Phantom for MRI and CT Reconstruction Pipelines: Beyond Shepp-Logan

Name of the presenter

Hakkel, Tamás

Institute/workplace of the presenter

Semmelweis University, Department of Biophysics and Radiation Biology, and Mediso Ltd.

Authors

Tamás Hakkel¹, Noémi Kovács², József Sinkó³, Máthé Domokos², Szigeti Krisztián²
1: Semmelweis University, Department of Biophysics and Radiation Biology and Mediso Ltd.
2: Semmelweis University, Department of Biophysics and Radiation Biology
3: Mediso Ltd.

Text of the abstract

Introduction: Developing new medical imaging methods relies heavily on simulations. While simple static models like the Shepp-Logan phantom are common, and complex proprietary 4D models exist, there is a lack of accessible, intermediate dynamic solutions.
Aims: We aimed to bridge the gap between simplistic static phantoms and sophisticated 4D models. We developed a customizable, open-source dynamic geometric phantom to facilitate testing and support the quantitative evaluation of dynamic MRI and CT pipelines.
Methods: The phantom utilizes constructive solid geometry, combining superellipsoids to schematically model a human torso, including lungs, a four-chamber heart, major vessels, the liver, the stomach, and the skeleton. This allows high-speed rendering on arbitrary grids and full parameterization of voxel intensities and organ volumes. Respiratory and cardiac motions are modeled via built-in signal generators. A high-performance reference implementation was written, which is available as a library and a standalone console application.
Result: Validation across four experiments demonstrated the phantom's capabilities. Volumetric validation showed excellent agreement between input and measured volumes: lung volumes exhibited a maximum relative error of 5.5% (median 2.5%), while cardiac chamber volume errors ranged from 1.76% to 3.67%. Benchmarking confirmed high-speed rendering performance across various grid sizes. Additionally, a simulated Cartesian MRI acquisition successfully demonstrated the phantom's utility in quantifying motion-induced artifacts.
Conclusion: This dynamic geometric phantom serves as an effective intermediate simulation tool. Providing full control over spatial and temporal resolution, the tissue intensities, and organ dynamics, it integrates seamlessly into imaging research pipelines.
Funding: Supported by the Hungarian National Research, Development and Innovation Office (EKÖP-KDP 2024), the EU Artificial Intelligence National Laboratory (MILAB, RRF-2.3.1-21-2022-00004), and the Horizon 2020 Program (739593: HCEMM).