Radiomic Feature Robustness and Reproducibility in Quantitative Bone Radiography: A Study on Radiologic Parameter...

Area of Science:

• Medical Imaging
• Radiology
• Quantitative Imaging

Background:

• Radiomic features are increasingly used in medical imaging for quantitative analysis.
• The robustness of these features to variations in imaging parameters is critical for clinical translation.
• Understanding parameter influence is essential for reproducible radiomic studies.

Purpose of the Study:

• To assess the impact of various radiologic parameters on the robustness of radiomic features.
• To identify radiomic features that are stable across different imaging settings.
• To evaluate the reproducibility of radiomic features using Bland-Altman analysis.

Main Methods:

• A tibia bone phantom was imaged using varied kV, mAs, filtration, tube angles, and source skin distance (SSD).
• Radiomic features from histogram, gradient, run-length matrix, co-occurrence matrix, autoregressive model, and wavelet sets were extracted.
• Feature robustness was quantified using coefficient of variation (COV), and reproducibility was assessed with Bland-Altman analysis.

Main Results:

• 22%, 34%, and 45% of features were robust (COV ≤ 5%) against kV, mAs, and SSD, respectively.
• No features were robust against filtration or tube angle; 100% and 76% showed high variation (COV > 20%).
• Co-occurrence matrix and histogram features (e.g., sum-average, sum-entropy, correlation, mean, percentiles) demonstrated the highest robustness.

Conclusions:

• Radiologic parameters significantly influence radiomic feature values, necessitating careful consideration in quantitative bone studies.
• Robust features with low COV are recommended for clinical and research applications.
• Bland-Altman analysis can identify reproducible radiomic features for consistent quantitative analysis.