Feature Analysis of Hyperpolarized Helium-3 Pulmonary MRI in Asthmatics versus Non-Asthmatics

Computational techniques for parenchymal characterization using CT have demonstrated significant research potential [1, 2]. We describe similar quantitative analysis of the lung using hyperpolarized helium-3 MR ventilation image data. This study consisted of a total of 55 subjects (47 asthmatics and 8 non-asthmatics). Each subject was imaged both before and after respiratory challenge. Additionally, each subject underwent a complete pulmonary function exam. Following image acquisition, approximately 1600 statistical features were calculated from the segmented lungs (and sub-regions) in each image. Each of these features were ranked along with the 27 pulmonary function test (PFT) values using a mutual information based feature subset selection algorithm. It is shown that several image features perform much better in characterizing clinical diagnosis compared with the current clinical gold-standard PFT values.