Radiomic features of breast parenchyma: assessing differences between FOR PROCESSING and FOR PRESENTATION digital...

Area of Science:

• Medical Imaging Analysis
• Radiomics and Quantitative Imaging
• Breast Cancer Screening Technologies

Background:

• Full-field digital mammography (FFDM) is a standard for breast cancer screening.
• Radiomics extracts quantitative features from medical images, offering insights beyond visual assessment.
• Understanding data processing differences (FOR PROCESSING vs. FOR PRESENTATION) is crucial for consistent radiomics analysis.

Purpose of the Study:

• To compare radiomics features extracted from FFDM images processed differently (FOR PROCESSING vs. FOR PRESENTATION).
• To evaluate the similarity and differences in segmentation and textural features between the two FFDM data types.
• To determine the suitability of each data type for assessing breast density using radiomics.

Main Methods:

• 165 women underwent FFDM; breasts were segmented into dense and non-dense areas.
• Radiomics features (First Order, GLCM, GLRLM, GLSZM) were computed from both FOR PROCESSING and FOR PRESENTATION images.
• Statistical analyses included Bland-Altman, Dice, kappa, Wilcoxon, Spearman, and intraclass correlation; bilateral symmetry and percent density (PD) were assessed.

Main Results:

• Segmentation differed significantly between FOR PROCESSING and FOR PRESENTATION images.
• Bilateral symmetry was higher for features derived from FOR PROCESSING data.
• While many features correlated, specific textural features showed moderate correlations with percent density, varying by data type and breast region.

Conclusions:

• Radiomics texture features from FOR PROCESSING FFDM images appear more appropriate for breast density assessment.
• Differences in image processing impact segmentation and feature analysis, highlighting the importance of data standardization.
• Further research may refine radiomics applications for improved mammographic analysis.