Related Experiment Video
Updated: Aug 5, 2026
10:17
Guidelines and Experience Using Imaging Biomarker Explorer (IBEX) for Radiomics
Published on: January 8, 2018
Inter-Observer Reproducibility of [18F]FDG PET/CT Radiomic Features in Primary Breast Carcinoma
Alexandru Mitoi1, Raluca Mititelu2,3, Cosmin Medar4,5
1Doctoral Program Studies, University of Medicine and Pharmacy "Carol Davila", 050474 Bucharest, Romania.
Journal of Imaging
|July 27, 2026
Summary
Radiomic feature stability in [18F]FDG PET/CT for breast cancer is crucial for clinical use. Most features showed excellent reproducibility, but shape features were less stable, informing future predictive radiomics studies.
Area of Science:
- Medical Imaging
- Radiomics
- Oncology
Background:
- Radiomic feature stability is vital for clinical translation of [18F]FDG PET/CT in breast carcinoma.
- Inter-observer segmentation variability's impact on radiomic features is not well understood.
Purpose of the Study:
- To evaluate inter-observer reproducibility of radiomic features in [18F]FDG PET/CT for breast cancer.
- To establish a feature reduction strategy for stable, non-redundant radiomic features.
Main Methods:
- Extracted 107 radiomic features from [18F]FDG PET/CT scans of 42 breast cancer patients.
- Used an IBSI-aligned PyRadiomics workflow with two independent segmentations and manual correction.
- Quantified reproducibility using ICC(A,1) and applied a two-stage reduction strategy.
Main Results:
- Good segmentation agreement (mean Dice 0.847).
- Most features (75.7%) demonstrated excellent reproducibility (ICC ≥ 0.90), median ICC 0.972.
- Shape features based on maximum lesion extension showed poor reproducibility (ICC 0.10-0.25).
Conclusions:
- A methodological framework for assessing radiomic feature stability in breast cancer was proposed.
- The study identified 19 stable, non-redundant features and classified feature stability.
- Results support the use of stable radiomic features for future predictive PET radiomics in breast carcinoma.
