Quantitative computed tomography imaging-based clustering differentiates asthmatic subgroups with distinctive clinical phenotypes

Publication date: Available online 29 January 2017
Source:Journal of Allergy and Clinical Immunology
Author(s): Sanghun Choi, Eric A. Hoffman, Sally E. Wenzel, Mario Castro, Sean Fain, Nizar Jarjour, Mark L. Schiebler, Kun Chen, Ching-Long Lin
BackgroundImaging variables including airway diameter, wall thickness and air-trapping have been found to be important metrics when differentiating severe asthmatics from nonsevere asthmatics and healthy subjects.ObjectiveThe objective of this study was to identify imaging-based clusters and to explore the association of the clusters with existing clinical metrics.MethodsWe performed an imaging-based cluster analysis using quantitative computed tomography-based structural and functional variables extracted from the respective inspiration and expiration scans of 248 asthmatics. The imaging-based metrics included a broader-set of multiscale variables such as inspiratory airway dimension, expiratory air-trapping and registration-based lung deformation (inspiration vs. expiration). Asthma subgroups derived from a clustering method were associated with subject demography, questionnaire, medication history, and biomarker variables.ResultsCluster 1 patients were early-onset younger nonsevere asthmatics with reversible airflow obstruction, who showed normal airway structure; Cluster 2 patients were a mix of nonsevere and severe asthmatics with marginal inflammation, who exhibited airway luminal narrowing without wall thickening. Cluster 3 and 4 patients were dominated by severe asthmatics. Cluster 3 patients were obese females with reversible airflow obstruction who exhibited airway wall thickening without airway narrowing. Cluster 4 patients were late-onset older males with persistent airflow obstruction, exhibiting significant air-trapping and reduced regional deformation. Clusters 3 and 4 patients also showed decreased lymphocyte and increased neutrophils, respectively.ConclusionsFour image-based clusters were identified and shown to be correlated with clinical characteristics. Such clustering serves to differentiate asthma subgroups which may be used as a basis for the development of new therapies.

Teaser

We identified four asthma subgroups using a cluster analysis composed of imaging variables, which were associated with clinical metrics. Identifying imaging-based clusters could enable practical cluster-based therapeutic interventions.


http://ift.tt/2khBxsk