Related Experiment Videos
[Classification of spot-shaped lung changes by texture analysis]
J F Desaga1, J Dengler, P Ipolt
1Röntgenabteilung Innere Medizin des Medizinischen Zentrums für Radiologie des Klinikums der Universität, Giessen.
Rofo : Fortschritte Auf Dem Gebiete Der Rontgenstrahlen Und Der Nuklearmedizin
|December 1, 1987
Summary
This study used textural analysis of chest x-rays to classify pneumoconiosis opacities. The method accurately identified known classes and showed promise for classifying new cases.
Area of Science:
- Radiology
- Medical Imaging
- Computational Pathology
Background:
- Pneumoconiosis diagnosis relies on identifying opacities in chest X-rays.
- Accurate classification of pneumoconiosis subtypes is crucial for patient management.
Purpose of the Study:
- To develop and evaluate a texture analysis method for classifying opacities in pneumoconiosis from digitized chest X-rays.
- To assess the accuracy of the proposed classification method on both training and independent test datasets.
Main Methods:
- Utilized a set of 10 texture parameters derived from algorithms including edge detection, local extremes, difference statistics, co-occurrence matrix, and power spectrum analysis.
- Applied these parameters to digitized chest X-ray images for classification.
- Evaluated performance on a training set and a separate test set with novel classes.
Main Results:
- Achieved high classification accuracy of 99% for known classes within the training set.
- Demonstrated a classification accuracy of 82% for a test set containing classes not previously encountered.
- The 10 selected texture parameters provided good discrimination between different classes of pneumoconiosis opacities.
Conclusions:
- Texture analysis of digitized chest X-rays is a highly effective method for classifying pneumoconiosis opacities.
- The developed algorithm shows strong potential for accurate diagnosis and classification of pneumoconiosis, even for previously unseen patterns.