Related Experiment Videos
Determining the view of chest radiographs
Thomas M Lehmann1, O Güld, Daniel Keysers
1Department of Medical Informatics, Aachen University of Technology (RWTH), Pauwelsstrasse 30, 52057 Aachen, Germany. lehmann@computer.org
Journal of Digital Imaging
|December 12, 2003
Summary
This study introduces an automated method for distinguishing frontal from lateral chest X-rays, achieving 99.7% accuracy. This technique enhances medical imaging analysis and computer-assisted diagnosis systems.
Area of Science:
- Medical Imaging
- Radiography Analysis
- Computer-Assisted Diagnosis
Background:
- Accurate classification of chest radiograph views (frontal vs. lateral) is crucial for medical imaging systems.
- Existing methods require sophisticated approaches, posing challenges for implementation.
Purpose of the Study:
- To develop an automated, accurate, and easily implementable algorithm for classifying chest radiograph views.
- To improve preprocessing for computer-assisted diagnosis and image retrieval systems.
Main Methods:
- Radiographs were downsized, and distance measures were applied for nearest-neighbor classification.
- Experiments utilized 1,867 clinical routine radiographs with a 5-nearest-neighbor scheme.
- Evaluated feature images of 32x32 and 8x8 pixels, using tangent distance and normalized cross-correlation.
Main Results:
- The best correctness achieved was 99.7% using 32x32 feature images, tangent distance, and 5-nearest-neighbor classification.
- Normalized cross-correlation yielded 99.6% and 99.3% correctness for 32x32 and 8x8 feature images, respectively.
- Errors were primarily due to pathologies, metal artifacts, or non-routine conditions.
Conclusions:
- The proposed algorithm offers superior performance compared to existing methods.
- The approach is highly accurate, robust, and simple to implement for routine clinical use.
- This method enhances the efficiency and reliability of automated medical image analysis.