Angular relational signature-based chest radiograph image view classification.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Machine Learning

Background:

• Computer-aided diagnosis (CAD) systems for chest X-ray (CXR) screening require accurate image view identification.
• Distinguishing between frontal and lateral CXR views is crucial for effective screening as different techniques may apply to each view.

Purpose of the Study:

• To develop and validate a novel, fully automatic technique for classifying frontal and lateral CXR images.
• To enhance the efficiency and accuracy of CXR screening processes within CAD systems.

Main Methods:

• Feature extraction using an "angular relational signature" derived from a force histogram.
• Classification using three state-of-the-art algorithms: multi-layer perceptron, random forest, and support vector machine.
• Validation on a large dataset of 8100 CXR images from the U.S. National Library of Medicine (NLM).

Main Results:

• Achieved classification accuracy close to 100% for distinguishing frontal and lateral CXR views.
• Demonstrated superior processing time, completing the classification of the entire test dataset in approximately 2 seconds.
• The developed method's accuracy is comparable to existing state-of-the-art techniques.

Conclusions:

• The novel technique for classifying CXR image views is highly accurate and efficient.
• The method's speed and accuracy make it a practical and valuable tool for computer-aided diagnosis in medical imaging.
• This approach offers a significant advancement for automated CXR screening systems.