Related Experiment Video
Updated: Jun 26, 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
Texture classification-based segmentation of lung affected by interstitial pneumonia in high-resolution CT
Panayiotis Korfiatis1, Christina Kalogeropoulou, Anna Karahaliou
1Department of Medical Physics, School of Medicine, University of Patras, Patras, Greece.
Medical Physics
|January 30, 2009
Summary
This study presents an automated lung field segmentation algorithm for high-resolution computed tomography (HRCT) that accurately identifies interstitial pneumonia (IP) patterns. The method achieves high accuracy, comparable to human observers, for computer-aided diagnosis (CAD) systems.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Radiology
Background:
- Accurate lung field (LF) segmentation in high-resolution computed tomography (HRCT) is crucial for computer-aided diagnosis (CAD) but is challenged by pathologies like interstitial pneumonia (IP).
- Existing segmentation methods struggle with the complex lung borders caused by various IP patterns, impacting diagnostic accuracy.
Purpose of the Study:
- To develop and evaluate a novel 2D LF segmentation algorithm specifically designed for HRCT images affected by interstitial pneumonia (IP).
- To compare the performance of the proposed algorithm against a traditional gray level thresholding (GLT) method and an alternative feature set.
Main Methods:
- The algorithm utilizes k-means clustering and a filling operation for initial LF estimation, followed by iterative support vector machine (SVM) neighborhood labeling.
- Two feature sets were evaluated: gray level and wavelet coefficient statistics, and gray level averaging and gradient features.
- Segmentation accuracy was assessed using area overlap and shape differentiation metrics (d(mean), d(rms), d(max)) against manual tracings on 22 HRCT cases with diverse IP patterns.
Main Results:
- The proposed method using gray level and wavelet coefficient statistics achieved the highest segmentation accuracy (overlap=0.954, d(mean)=1.080 mm), outperforming the alternative feature set and the GLT-based method.
- Segmentation accuracy decreased with increasing IP severity but remained statistically comparable across mild, moderate, and severe cases.
- The developed algorithm's accuracy was found to be within the range of interobserver variability.
Conclusions:
- The proposed LF segmentation algorithm, leveraging gray level and wavelet coefficient statistics, demonstrates superior accuracy for HRCT images with interstitial pneumonia.
- This method offers a robust solution for automated lung segmentation, potentially serving as a foundational step for advanced CAD schemes in IP detection and analysis.
- The algorithm's performance is reliable across different IP severity levels, making it a valuable tool in clinical radiology.
