Related Experiment Video
Updated: Oct 11, 2025
10:17
Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics
Published on: January 8, 2018
13.3K
Reproducibility and Repeatability of CBCT-Derived Radiomics Features.
Hao Wang1,2,3, Yongkang Zhou4, Xiao Wang2
1Department of Radiation Oncology, Shanghai Chest Hospital, Shanghai Jiaotong University, Shanghai, China.
Frontiers in Oncology
|December 6, 2021
Summary
This study evaluated the reproducibility and repeatability of radiomics features from cone-beam computed tomography (CBCT) in head and neck and pelvic cancer patients. Results showed feature stability varies by type, site, and time, impacting clinical use.
Area of Science:
- Medical Imaging and Radiomics
- Oncology
- Quantitative Imaging
Background:
- Radiomics extracts quantitative features from medical images, offering potential for cancer diagnosis and treatment response assessment.
- Cone-beam computed tomography (CBCT) is increasingly used in radiation therapy, making its radiomics features relevant for clinical applications.
- Assessing the reliability of CBCT radiomics features is crucial before their integration into clinical workflows.
Purpose of the Study:
- To determine the reproducibility and repeatability of radiomics features extracted from cone-beam computed tomography (CBCT) images.
- To compare the performance of two radiomics software packages in feature extraction.
- To evaluate feature stability across different anatomical sites (head and neck, pelvis) and over time.
Main Methods:
- Retrospective collection of CBCT images from 20 cancer patients (10 head and neck, 10 pelvic) acquired on different days.
- Extraction of 18 radiomics features using two distinct software packages.
- Assessment of reproducibility using intraclass correlation coefficient (ICC) and repeatability using concordance correlation coefficient (CCC).
Main Results:
- First-order histogram-based features (skewness, entropy) showed good reproducibility (ICC ≥ 0.80).
- GLCM-based features demonstrated good agreement between software packages, except for GLCM-correction.
- Repeatability (CCC) for most first-order and GLCM features exceeded 0.80 over 2 days but decreased after 5 days; two GLRLM features showed high volume correlation (R² ≥ 0.8).
Conclusions:
- This is the first study to compare CBCT radiomics feature reproducibility and repeatability in both head and neck and pelvic cancer sites.
- Feature stability is influenced by the feature type, anatomical site, and the time interval between scans.
- The findings highlight the need for site- and protocol-specific validation of CBCT radiomics features for clinical application.
