Assessing Agreement between Radiomic Features Computed for Multiple CT Imaging Settings.

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Radiomics uses quantitative image features (QIFs) to analyze tumor characteristics.
• Radiological images are often acquired using diverse equipment and settings, potentially impacting QIF consistency.

Purpose of the Study:

• To evaluate the impact of varying slice thickness and reconstruction algorithms on the agreement of QIFs from CT images.
• To assess inter-setting variability in radiomic feature extraction for lung cancer.

Main Methods:

• CT scans from 32 lung cancer patients were used.
• Raw data were reconstructed into six series using two algorithms (Lung/Standard) and three slice thicknesses (1.25mm, 2.5mm, 5mm).
• 89 QIFs were computed, clustered to reduce redundancy, and agreement was assessed using concordance correlation coefficients (CCCs).

Main Results:

• Twenty-three non-redundant feature groups were identified.
• Best agreement (CCCs>0.8) was observed between similar settings (e.g., 1.25S vs 2.5S).
• Eight feature groups (size, shape, texture) showed high agreement (average CCC>0.8) across settings.

Conclusions:

• Significant inter-setting disagreements in QIFs arise from variations in CT reconstruction algorithms and slice thickness.
• Standardizing imaging acquisition protocols is essential for reliable radiomics studies and consistent tumor phenotype characterization.