Non-uniform central airways ventilation model based on vascular segmentation

Algorithm of Pulmonary Vascular Segment and Centerline Extraction

Because pulmonary vascular lesions are harmful to the human body and difficult to detect, computer-assisted diagnosis of pulmonary blood vessels has become the focus and difficulty of the current research. An algorithm of pulmonary vascular segment and centerline extraction which is consistent with the physician's diagnosis process is proposed for the first time. We construct the projection of maximum density, restore the vascular space information, and correct random walk algorithm to satisfy automatic and accurate segmentation of blood vessels. Construct a local 3D model to restrain Hessian matrix when extracting centerline. In order to assist the physician to make a correct diagnosis and verify the effectiveness of the algorithm, we proposed a visual expansion model. According to the 420 high-resolution CT data of lung blood vessels labeled by physicians, the accuracy of segmentation algorithm AOM reached 93%, and the processing speed was 0.05 s/frame, which achieved the clinical application standards.

Effect of static vs. dynamic imaging on particle transport in CT-based numerical models of human central airways

Advances in quantitative computed tomography (CT) has provided methods to assess the detailed structure of the pulmonary airways and parenchyma, providing the means of applying computational fluid dynamics-based modeling to better understand subject-specific differences in structure-to-function relationships. Most of the previous numerical studies, seeking to predict patterns of inhaled particle deposition, have considered airway geometry and regional ventilation derived from static images. Because geometric alterations of the airway and parenchyma associated with regional ventilation may greatly affect particle transport, we have sought to investigate the effect of rigid vs. deforming airways, linear vs. nonlinear airway deformations, and step-wise static vs. dynamic imaging on particle deposition with varying numbers of intermediate lung volume increments. Airway geometry and regional ventilation at different time points were defined by four-dimensional (space and time) dynamic or static CT images. Laminar, transitional, and turbulent air flows were reproduced with a three-dimensional eddy-resolving computational fluid dynamics model. Finally, trajectories of particles were computed with the Lagrangian tracking algorithm. The results demonstrated that static-imaging-based models can contribute 7% uncertainty to overall particle distribution and deposition primarily due to regional flow rate (ventilation) differences as opposed to geometric alterations. The effect of rigid vs. deforming airways on serial distribution of particles over generations was significantly smaller than reported in a previous study that used the symmetric Weibel geometric model with smaller flow rate. Rigid vs. deforming airways were also shown to affect parallel particle distribution over lobes by 8% and the differences associated with use of static vs. dynamic imaging was 18%. These differences demonstrate that estimates derived from static vs. dynamic imaging can significantly affect the assessment of particle distribution heterogeneity. The effect of linear vs. nonlinear airway deformations was within the uncertainty due to mesh size.

Modeling the Nonlinear Motion of the Rat Central Airways

Advances in volumetric medical imaging techniques allowed the subject-specific modeling of the bronchial flow through the first few generations of the central airways using computational fluid dynamics (CFD). However, a reliable CFD prediction of the bronchial flow requires modeling of the inhomogeneous deformation of the central airways during breathing. This paper addresses this issue by introducing two models of the central airways motion. The first model utilizes a node-to-node mapping between the discretized geometries of the central airways generated from a number of successive computed tomography (CT) images acquired dynamically (without breath hold) over the breathing cycle of two Sprague-Dawley rats. The second model uses a node-to-node mapping between only two discretized airway geometries generated from the CT images acquired at end-exhale and at end-inhale along with the ventilator measurement of the lung volume change. The advantage of this second model is that it uses just one pair of CT images, which more readily complies with the radiation dosage restrictions for humans. Three-dimensional computer aided design geometries of the central airways generated from the dynamic-CT images were used as benchmarks to validate the output from the two models at sampled time-points over the breathing cycle. The central airway geometries deformed by the first model showed good agreement to the benchmark geometries within a tolerance of 4%. The central airway geometry deformed by the second model better approximated the benchmark geometries than previous approaches that used a linear or harmonic motion model.