Reference standard and statistical model for intersite and temporal comparisons of CT attenuation in a multicenter...

Area of Science:

• Medical Imaging Physics
• Radiology
• Pulmonary Medicine

Background:

• Multicenter studies require consistent data acquisition.
• Variability in computed tomography (CT) scanner performance can impact research findings.
• The COPDGene study collects crucial pulmonary CT data for genetic epidemiology.

Purpose of the Study:

• To detect and analyze anomalies in CT scanner performance over time.
• To ensure CT attenuation measurement accuracy for the COPDGene study.
• To establish a method for tracking CT scanner consistency across multiple sites.

Main Methods:

• Development of a custom CT reference standard "Test Object" with materials simulating lung tissue, air, water, and acrylic.
• Scanning of 19 Test Objects across 42 CT scanners at 20 sites over 17 months.
• Statistical analysis using a mixed linear model to assess influences of reconstruction kernels, tube current, scanner models, and temporal consistency.

Main Results:

• Significant differences in CT attenuation were observed based on reconstruction kernels, tube current, CT scanner models, and temporal consistency.
• Lung equivalent foam and internal air materials showed significant variations between different scanner settings and models.
• No significant effects were found when comparing different examples of the Test Objects against reference scans, indicating inter-object consistency.

Conclusions:

• The developed CT Test Object effectively detects significant scanner-induced alterations in CT attenuation relevant to lung density measurements.
• This analysis method can identify CT scanner anomalies and statistically account for scanner variations and temporal changes in multicenter trials.
• Understanding and modeling individual scanner performance is crucial for maintaining data integrity in large-scale pulmonary CT studies.