Image dissimilarity-based quantification of lung disease from CT

Area of Science:

• Medical Imaging Analysis
• Computer-Aided Diagnosis
• Pulmonary Disease Quantification

Background:

• Accurate classification of medical images is crucial for computer-aided diagnosis.
• Existing methods for quantifying pulmonary diseases like COPD often have limitations.

Purpose of the Study:

• To propose and evaluate a new medical image classification method based on region dissimilarities.
• To apply this method for quantifying chronic obstructive pulmonary disease (COPD) in computed tomography (CT) images.

Main Methods:

• Images are mapped into a dissimilarity space using an image dissimilarity measure.
• A standard vector space-based classifier is applied within this dissimilarity space.
• The approach focuses on collections of regions of interest for classification.

Main Results:

• The proposed method achieved an area under the receiver operating characteristic curve of 0.817 for COPD quantification in CT images.
• This performance was significantly better than combining individual region classifications.
• It also surpassed common computerized quantitative measures in pulmonary CT analysis.

Conclusions:

• The proposed dissimilarity-based image classification approach offers improved performance for medical image analysis.
• This method shows significant potential for computer-aided diagnosis, particularly in quantifying pulmonary abnormalities like COPD.