Multi-institutional MRI-based radiomic pilot study to measure the variations between scanner vendors and imaging...

Area of Science:

• Medical Imaging
• Radiomics
• Quality Assurance

Background:

• Multi-institutional clinical trials rely on MRI imaging for treatment decisions.
• Variability in image quality across different MRI vendors and models poses challenges.
• A unifying quality control protocol is needed to ensure MRI image comparability in studies.

Purpose of the Study:

• To develop a 3D radiomic phantom for quantifying radiomic feature consistency across MRI scanners.
• To enable standardized quality assessment in multi-institutional MRI studies.

Main Methods:

• A prototype phantom with 3D-printed grid and egg-shaped objects was created using Polylactic Acid (PLA).
• Objects were scanned using T1- and T2-weighted sequences on multiple MRI scanners (Philips, GE, Siemens, ViewRay) across institutions.
• 107 radiomic features were extracted, and Coefficients of Variation (CV) were analyzed to compare consistency.

Main Results:

• Radiomic feature consistency varied by object shape, with grid objects showing higher reproducibility than egg-shaped objects.
• GE scanners exhibited better reproducibility; ViewRay scanners showed consistency with the TrueFISP sequence (median CV <10%).
• T1w sequences demonstrated slightly better reproducibility than T2w sequences.

Conclusions:

• Significant intra-scanner variations in radiomic features were observed.
• A standardized radiomic phantom is essential for characterizing MRI scanners and sequences.
• Establishing a baseline for radiomic features is crucial for multi-institutional MRI studies.