Creating an anthropomorphic digital MR phantom--an extensible tool for comparing and evaluating quantitative imaging...

Area of Science:

• Medical Imaging
• Quantitative Analysis
• Computational Phantoms

Background:

• Assessing bias and variance in image quantification algorithms is crucial for clinical applications.
• Current assessment methods use grid-based digital reference objects (DROs) or digital anthropomorphic phantoms.
• Publicly available phantoms can generate realistic DROs with pathological heterogeneity.

Purpose of the Study:

• To develop a methodology for generating model-based anthropomorphic DROs representing realistic, heterogeneous pathology.
• To assess the impact of spatial smoothing on quantitative imaging algorithm parameters.

Main Methods:

• A publicly available vascular input function and digital brain phantom were used.
• A general kinetic model (GKM) was employed to synthesize a dynamic contrast-enhanced (DCE) MRI exam.
• Clinical DCE-MRI data were used to estimate GKM parameters and create a phantom with normal and tumor tissues.
• Voxel-wise Bland-Altman analysis assessed the effect of spatial smoothing.

Main Results:

• A methodology was established to generate DROs with realistic, enhancing pathology.
• Spatial smoothing with a small Gaussian kernel showed negligible differences in estimated GKM parameters.
• The developed DROs can be used to assess quantitative imaging algorithms.

Conclusions:

• An extensible methodology for creating model-based anthropomorphic DROs was reported.
• These DROs incorporate normal and pathological tissues for robust algorithm assessment.
• The approach enhances the reliability of quantitative imaging in clinical research and practice.