The application of statistical shape modeling for lung morphology in aerosol inhalation dosimetry

Abstract

Even with the advance in medical imaging techniques such as CT/MRI, it is still challenging and time-consuming to reconstruct anatomically accurate lung geometries. It is even more challenging to study variability in inhalation dosimetry or pulmonary drug delivery, which requires a large cohort of lung models to ensure statistically significant results. This study used the statistical shape modeling (SSM) that bases on a limited number of lung models (40) to generate infinitely large numbers of parameterized models, which can span all major features inherent in the database of lung geometries. We demonstrated this model in lung models with more than 400 outlets (G9), which first identified the principal components (PCs) of base models, and then regenerated new models by systematically varying the mode (eigenvector) and its eigenvalues. The new models included airway remodeling at varying locations (left upper lobe and right lower lobe) and with varying levels of airway distensibility (compliance) and constriction (resistance). Airflow and aerosol dynamics within these lung geometries were numerically computed and compared. Results showed that even though the airway remodeling can be local, its influences on flow partition and deposition distribution can be global. Asthma-induced bronchiolar constriction, when severe, can strikingly alter the airflow and particle deposition mapping throughout the lungs. The highest deposition variability due to airway remodeling was found to come from particles of 4–10 μm in the upper lobes, and of 10–20 μm in the lower lobe. Statistical shape modeling is an imaging processing method that has often been used in computer sciences. This is the first study, to the author's knowledge, that SSM was applied in lung models with high complexity to quantify the resultant variances from these geometry remodeling. This method was also applied to lung models with 3000 outlets (G11) to generate diseased lung models at varying locations.