Chronic obstructive pulmonary disease (COPD) is a leading cause of death worldwide. In this chapter, a novel computer-aided diagnosis (CAD) technique is introduced for automatic image-based COPD diagnosis and GOLD stage classification. The technique includes a highly accurate segmentation method that utilizes thoracic computed tomographic (CT) image data to estimate the lung air volumes over respiration cycle. The segmentation method was developed based on the biophysics of the lung tissue in conjunction with the lung morphometrics. The technique then exploits the lung air volume data obtained from the segmentation method to automatically extract 23 features pertaining to the lung air volume distribution and variation over inspiration to expiration phases. Relationships between the extracted features and spirometric data were investigated and indicated strong correlation. Furthermore, the discriminatory power of all features were examined using sequential forward selection and sequential backward selection algorithms. This led to the selection of 12 features with the most discriminatory power for training the classifiers. Central to this system is a classifier that was developed by analyzing CT images of a cohort of 69 subjects, including 13 normal and 56 COPD patients with known COPD GOLD stages. The performance of the classifiers was evaluated and compared utilizing leave-m-out cross-validation method with m = 7. Results obtained in this investigation showed the highest accuracy with the Naive Bayes classifier, where an overall accuracy of over 84% was obtained. This demonstrates the proposed CAD system's potential as a clinically viable image-based COPD diagnosis method that requires CT images only as input and does not require pulmonary function test measurements.