Three-dimensional printing of patient-specific computed tomography lung phantoms: a reader study

Area of Science:

• Medical Imaging
• Radiology
• Biomedical Engineering

Background:

• Heterogeneity in medical imaging data due to varying systems and protocols impedes reproducible quantitative analysis in clinical decision-support algorithms.
• Developing consistent ground-truth targets is crucial for validating the generalizability of these algorithms across diverse healthcare settings and imaging protocols.

Purpose of the Study:

• To investigate the use of patient-specific 3D-printed lung phantoms for providing consistent ground-truth targets.
• To assess the reliability and reproducibility of a novel 3D-printing method, PixelPrint, for creating lifelike computed tomography (CT) lung phantoms.

Main Methods:

• PixelPrint was utilized to 3D-print CT lung phantoms from COVID-19 patient data, controlling voxel-by-voxel density.
• A blinded reader study involved five radiologists comparing patient and phantom images for characteristics and diagnostic confidence.
• Linear mixed models were used to assess effect sizes, and PixelPrint's production reproducibility was evaluated.

Main Results:

• Images of patients and phantoms exhibited minimal variation in estimated mean values.
• The difference between phantom and patient images was within one-third of inter- and intra-reader variabilities, indicating high consistency.
• PixelPrint demonstrated high production repeatability, with phantoms printed from the same data showing greater similarity than clinical-dose acquisitions of a single phantom.

Conclusions:

• PixelPrint reliably produces lifelike CT lung phantoms.
• These phantoms can serve as valuable ground-truth targets for validating decision-support algorithms and optimizing imaging protocols.
• The developed phantoms facilitate the assessment of algorithm generalizability across different health centers and imaging protocols.