Related Experiment Video
Updated: Mar 11, 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
14.9K
Evaluation of Semi-automatic Segmentation Methods for Persistent Ground Glass Nodules on Thin-Section CT Scans
Young Jae Kim1, Seung Hyun Lee2, Chang Min Park3
1Biomedical Engineering Branch, Division of Precision Medicine and Cancer Informatics, Research Institute, National Cancer Center, Goyang, Korea.; Department of Plasma Bio Display, Kwangwoon University, Seoul, Korea.
Healthcare Informatics Research
|November 30, 2016
Summary
The level-set-based active contour model best segments persistent ground glass nodules (GGN) in CT scans. This method most closely matches radiologist segmentations, aiding computer-aided lung cancer diagnosis.
Area of Science:
- Medical Imaging Analysis
- Radiology
- Computer-Aided Diagnosis
Background:
- Accurate segmentation of persistent ground glass nodules (GGN) in thin-section CT images is crucial for lung cancer diagnosis.
- Various semi-automatic segmentation methods exist, but their comparative performance for GGNs needs evaluation.
Purpose of the Study:
- To compare the accuracy of five semi-automatic segmentation methods for persistent GGNs in thin-section CT images.
- To identify the most effective segmentation method for GGNs to support computer-aided diagnosis systems.
Main Methods:
- Five semi-automatic segmentation algorithms were applied: level-set-based active contour model, localized region-based active contour model, seeded region growing, K-means clustering, and fuzzy C-means clustering.
- Segmentation accuracy was quantified using the Dice coefficient, comparing algorithm results against manual segmentations by two radiologists.
Main Results:
- The level-set-based active contour model achieved the highest mean Dice coefficient (0.808), indicating the best similarity to manual segmentation.
- K-means clustering (0.7953) and fuzzy C-means clustering (0.7999) also showed strong performance.
- Seeded region growing yielded the lowest Dice coefficient (0.629).
Conclusions:
- The level-set-based active contour model is the most effective method for segmenting persistent GGNs in thin-section CT images.
- This algorithm's efficiency and similarity to manual segmentation are vital for developing advanced computer-aided diagnosis systems for lung cancer.
