Related Experiment Video
Updated: Mar 14, 2026
06:49
Orthotopic Implantation of Patient-Derived Cancer Cells in Mice Recapitulates Advanced Colorectal Cancer
Published on: February 10, 2023
2.9K
Reproducibility with repeat CT in radiomics study for rectal cancer
Panpan Hu1,2, Jiazhou Wang1,2, Haoyu Zhong1,2
1Department of Radiotherapy, Fudan University Shanghai Cancer Center, Shanghai, China.
Oncotarget
|September 28, 2016
Summary
Radiomics feature reproducibility in rectal cancer was assessed using repeated CT scans. Volume-normalized average slice features demonstrated the highest stability, crucial for reliable treatment monitoring and prognosis prediction.
Area of Science:
- Medical Imaging
- Oncology
- Radiomics
Background:
- Radiomics analysis involves extracting quantitative features from medical images.
- Assessing the reproducibility of these features is critical for their clinical application in cancer management.
Purpose of the Study:
- To evaluate the reproducibility of radiomics features in rectal cancer using repeated computed tomographic (CT) scans.
- To identify stable radiomics features for reliable rectal cancer analysis.
Main Methods:
- 40 stage II rectal cancer patients underwent two CT scans.
- 775 radiomics features were extracted and analyzed for reproducibility using concordance correlation coefficients (CCC) and inter-class correlation coefficients (ICC).
- Features were assessed based on original and volume-normalized values, with various extraction methods and LOG filters applied.
Main Results:
- Volume-normalized features exhibited significantly higher reproducibility than unnormalized features.
- Average value of all slices emerged as the most reproducible feature type.
- Among average type features, 496/775 showed high reproducibility (ICC ≥ 0.8), 225/775 medium (0.8 > ICC ≥ 0.5), and 54/775 low (ICC < 0.5).
- Image filtering had minimal impact on feature reproducibility.
Conclusions:
- Volume normalization is recommended for radiomics analysis in rectal cancer.
- Average type radiomics features are the most stable and reliable.
- These stable features hold promise for future applications in treatment monitoring and prognosis prediction for rectal cancer.
