Related Experiment Video
Updated: May 11, 2026
02:09
Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function
Published on: April 12, 2024
Lung texture in serial thoracic CT scans: registration-based methods to compare anatomically matched regions
Alexandra R Cunliffe1, Samuel G Armato, Xianhan M Fei
1Department of Radiology, The University of Chicago, Chicago, Illinois 60637, USA.
Medical Physics
|May 31, 2013
Summary
Comparing three demons registration methods for serial CT scans, calculating features on original scans minimized bias. This approach identifies reliable texture features for tracking disease progression without altering scan data.
Area of Science:
- Medical imaging analysis
- Radiomics and texture analysis
- Image registration techniques
Background:
- Serial computed tomography (CT) scans are crucial for monitoring disease progression.
- Accurate spatial matching of regions of interest (ROIs) between scans is essential for reliable texture analysis.
- Demons registration is a deformable image registration technique used for aligning medical images.
Purpose of the Study:
- To compare three demons registration-based methods for identifying spatially matched ROIs in serial CT scans.
- To evaluate the impact of different registration strategies on texture feature analysis.
- To identify optimal methods for texture analysis in serial CT imaging.
Main Methods:
- Retrospective collection of 27 patients' serial thoracic CT scans with no lung abnormalities.
- Placement of over 1000 ROIs in baseline scans, with anatomically matched ROIs in follow-up scans using three methods: original scan, resampled scan, and affine registration.
- Calculation of 140 texture features, followed by evaluation of feature value differences using Bland-Altman analysis to determine normalized bias and normalized range of agreement (nRoA).
Main Results:
- Significant differences in normalized bias were observed among the three registration methods.
- The lowest normalized bias (median: 0.06%) was achieved when texture features were calculated on original, non-deformed follow-up scans.
- A set of 20 texture features with low bias and variability (nRoA) were identified using the original scan method.
Conclusions:
- Calculating texture features on original serial CT scans, rather than deformed scans, minimizes bias in feature value changes.
- The demons registration method, when applied to original scans, facilitates reliable texture analysis for serial CT imaging.
- This approach enables accurate measurement of pathologic changes between CT scans without altering image data.
