Quality control of radiomic features using 3D-printed CT phantoms

Area of Science:

• Medical Imaging
• Radiomics
• 3D Printing Technology

Background:

• Quantitative radiomic measures lack standardization across computed tomography (CT) scans, hindering clinical applications.
• Radiomic feature robustness is crucial for reliable interpretation across different CT scan modes, protocols, software, and systems.

Purpose of the Study:

• To demonstrate the utility of custom-designed, 3D-printed phantoms with human-derived patterns for validating radiomic signature values.
• To assess the repeatability and reproducibility of radiomic features using a novel phantom in a clinical setting.

Main Methods:

• A multipurpose radiomic phantom was designed using a multimaterial, ultra-high-resolution 3D printer with voxel printing capabilities.
• Tissue regions of interest from CT scans of pancreas, lung, and liver tumors were incorporated into the phantom design.
• The phantom underwent 30 repeat CT scans using standard clinical protocols to evaluate feature repeatability and reproducibility across different scan modes.

Main Results:

• The 3D-printed phantom demonstrated structural similarity to patient DICOM images (structural similarity index of 0.71).
• Radiomic features exhibited low coefficients of variation ($$) across 30 repeat scans.
• Image noise, voxel size, and convolution kernels significantly impacted radiomic feature values, particularly gray level co-occurrence features.

Conclusions:

• Existing prognostic radiomic features show variability in practice and require cautious interpretation or exclusion from clinical use.
• Voxel-based 3D printing effectively reproduces CT-based tissue morphology.
• Custom 3D-printed phantoms provide a flexible and practical method for longitudinal validation and cross-system comparison of radiomic metrics.