Reported estimates of human airway dimensions are inconsistent across studies

Abstract

RationaleRespiratory diseases are a source of immense socioeconomic burden globally. In silico approaches can predict changes in human lung function due to disease or response to therapy. By stratifying patient-specific response a priori, these models can enable clinical-scale deployment of precision medicine strategies. Key to this is developing accurate organ geometries on which the models can be simulated. However, we lack analyses assessing the clinical applicability of reported airway dimension estimates. ObjectiveTo investigate physiologically-/anatomically-relevant airway dimension estimates and evaluate consistency across reported literature. MethodsWe conducted a systematic review of 37 published datasets. Airway wall thickness estimates were mined for healthy subjects and patients, and standardised to the Horsfield order airway generations. We simulated dynamic lung function to quantitatively assess their physiological relevance. We created an online database to make all datasets available to the research community. Measurements and Main ResultsReported human airway wall thickness estimates are inconsistent across studies. K-means clustering divided estimates for healthy subjects and patients into three and four clusters, respectively. Only one of the clusters in each category yielded anatomically-relevant estimates. Pressure-volume curves generated to assess physiological relevance also showed that only one cluster in each category exhibited plausible physiology. Principal Component Analysis weakly implicated imaging modalities to explain this inconsistency. ConclusionsReported airway dimension estimates are inconsistent and lack standardisation. To support future modelling efforts, we report physiologically-relevant estimates and introduce an open-access airway-dimension database to help standardise geometric inputs and quantify how measurement variability propagates to functional predictions.