Related Experiment Video
Updated: Nov 21, 2025
10:17
Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics
Published on: January 8, 2018
13.5K
Digital phantoms for characterizing inconsistencies among radiomics extraction toolboxes
Yushi Chang1,2, Kyle Lafata1, Chunhao Wang1
1Department of Radiation Oncology, Duke University Medical Center, Durham, United States of America.
Biomedical Physics & Engineering Express
|January 13, 2021
Summary
Digital phantoms reveal significant inconsistencies in radiomics extraction across different software toolboxes. Implementing quality assurance for radiomics extraction is crucial for reproducible and generalizable research findings.
Area of Science:
- Medical Imaging
- Radiomics
- Computational Biology
Background:
- Radiomics analysis involves extracting quantitative features from medical images.
- Variability in radiomics extraction methods can impact study reproducibility and generalizability.
- Standardized methods for assessing radiomics tool performance are needed.
Purpose of the Study:
- To develop and utilize digital phantoms for characterizing inconsistencies among radiomics extraction methods.
- To compare the performance of three radiomics toolboxes: CERR, IBEX, and an in-house platform.
- To identify sources of discrepancies in radiomics feature extraction.
Main Methods:
- Developed digital bar phantoms for intensity and texture features.
- Created heteromorphic sphere phantoms for shape feature analysis.
- Compared 61 radiomics features across CERR, IBEX, and an in-house toolbox.
- Quantified feature inconsistencies using concordance correlation coefficient (CCC) and Pearson correlation coefficient (PCC).
Main Results:
- Intensity and texture features showed limited perfect reproducibility (CCC=1) across toolboxes.
- Shape features like surface area exhibited inconsistencies due to varying mathematical definitions and implementations.
- Discrepancies were attributed to differences in pre-processing and calculation methods.
Conclusions:
- Developed digital phantoms effectively identify inconsistencies in radiomics extraction toolboxes.
- The findings underscore the critical need for quality assurance (QA) in radiomics extraction.
- Digital phantoms serve as valuable tools for ensuring the reproducibility and generalizability of radiomic studies.
