A computational framework is described that was developed for quantitative analysis of hyperpolarized helium-3 MR lung ventilation image data. This computational framework was applied to a study consisting of 55 subjects (47 asthmatic and eight normal). Each subject was imaged before and after respiratory challenge and also underwent spirometry. Approximately 1600 image features were calculated from the lungs in each image. Both the image and 27 spirometric features were ranked based on their ability to characterize clinical diagnosis using a mutual information-based feature subset selection algorithm. It was found that the top image features perform much better compared with the current clinical gold-standard spirometric values when considered individually. Interestingly, it was also found that spirometric values are relatively orthogonal to these image feature values in terms of informational content.