Related Experiment Video
Updated: Apr 27, 2026
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
13.4K
Test-retest reproducibility analysis of lung CT image features
Yoganand Balagurunathan1, Virendra Kumar, Yuhua Gu
1Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, USA.
Journal of Digital Imaging
|July 4, 2014
Summary
Quantitative CT image features show promise as prognostic biomarkers for non-small cell lung cancer (NSCLC). Reproducible size and texture features accurately predict patient outcomes, aiding in personalized treatment strategies for lung cancer.
Area of Science:
- Radiology and Oncology
- Medical Imaging Analysis
- Biomarker Discovery
Background:
- Quantitative imaging features from computed tomographic (CT) scans are potential biomarkers for non-small cell lung cancer (NSCLC).
- Reproducibility, non-redundancy, and dynamic range are crucial for feature utility as biomarkers.
- Developing robust quantitative features is essential for prognostic applications in NSCLC.
Purpose of the Study:
- To develop and validate reproducible, non-redundant quantitative three-dimensional (3D) CT features for NSCLC.
- To assess the prognostic utility of these selected image features in predicting patient outcomes.
- To identify informative imaging biomarkers for NSCLC prognosis.
Main Methods:
- Extracted 219 quantitative 3D features (size, shape, texture) from segmented NSCLC lesions in CT scans of 32 patients.
- Evaluated feature reproducibility using concordance correlation coefficient (CCC) across test-retest scans.
- Selected non-redundant features based on CCC, biological range, and feature independence measures.
Main Results:
- Identified 66 features (30.14%) with CCC ≥ 0.90 and acceptable dynamic range.
- Reduced to 42 non-redundant features using R² threshold.
- Selected features demonstrated predictive ability for radiological prognosis, with AUCs of 91% (size) and 92% (texture).
- Validated on an independent cohort, texture features achieved AUC of 0.84, outperforming conventional size-based features (AUC 0.80).
Conclusions:
- Identified reproducible and non-redundant quantitative CT image features with prognostic value in NSCLC.
- These features exhibit high intra-patient reproducibility and inter-patient biological range.
- Quantitative image features serve as informative and prognostic biomarkers for NSCLC, supporting clinical decision-making.
