Related Experiment Videos
Automated lung segmentation in digital lateral chest radiographs
S G Armato1, M L Giger, K Ashizawa
1Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago, Illinois 60637, USA.
Medical Physics
|September 2, 1998
Summary
We developed an automated method to segment lung fields in lateral chest X-rays, improving computer-aided diagnosis. This technique accurately identifies lung borders in lateral views, a crucial step for advanced medical imaging analysis.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Radiography
Background:
- Existing automated lung segmentation methods primarily focus on posteroanterior (PA) chest radiographs.
- Lateral chest radiographs contain vital diagnostic information routinely used by radiologists.
- Accurate lung segmentation in lateral views is essential for developing advanced computer-aided diagnostic (CAD) schemes.
Purpose of the Study:
- To develop a fully automated computerized scheme for segmenting lung fields in digital lateral chest radiographs.
- To enable computer analysis of lateral chest images, potentially enhancing current CAD systems.
Main Methods:
- An automated scheme employing pixel elimination, global gray-level histogram analysis, and iterative thresholding.
- Adaptive local gray-level thresholding refines contours based on anatomical regions.
- Segmentation is completed using smoothing and polynomial curve fitting.
Main Results:
- The automated scheme was tested on a database of 100 normal and 100 abnormal lateral chest radiographs.
- Quantitative comparison showed high agreement between computer-segmented and radiologist-delineated lung regions.
- 83% of normal and 84% of abnormal images achieved segmentation contours within three standard deviations of the mean inter-radiologist overlap.
Conclusions:
- The developed automated scheme provides accurate lung field segmentation in lateral chest radiographs.
- This method has the potential to significantly contribute to the development of advanced CAD systems utilizing lateral radiographic views.