Stability assessment of first order statistics features computed on ADC maps in soft-tissue sarcoma

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Radiomics quantifies medical images for detailed characterization.
• Assessing the stability and relevance of radiomic features is crucial for reliable analysis.
• Soft-tissue sarcomas (STSs) require robust imaging biomarkers for diagnosis and management.

Purpose of the Study:

• To evaluate the stability and relevance of radiomic features derived from diffusion-weighted magnetic resonance imaging (DW-MRI) in soft-tissue sarcomas.
• To establish criteria for feature acceptance based on transformations.
• To identify a stable subset of radiomic features for potential use in STSs.

Main Methods:

• Computed apparent diffusion coefficient (ADC) maps from DW-MRI scans of 18 STS patients.
• Extracted 37 intensity-based radiomic features from regions of interest (ROIs).
• Assessed feature stability by applying translations and rotations to ROIs and calculating intra-class correlation coefficients (ICCs).

Main Results:

• 31 out of 37 radiomic features met the acceptance criteria for stability (ICC > 0.6 after minimum transformation, < 0.4 after maximum translation).
• The methodology demonstrated a quantifiable approach to feature stability assessment.
• A significant portion of the radiomic features exhibited robustness to image transformations.

Conclusions:

• Radiomic feature stability analysis is a viable first step in feature selection for STSs.
• The findings support the use of stable radiomic features in quantitative medical imaging.
• This approach can enhance the reliability of radiomics in oncological applications.