**4. Challenges and future directions**

Challenges faced by the current CFPD models and numerical studies for unconventional lung aerosol dynamics are mostly induced by the point-mass and point-force assumptions, the resolution of medical imaging techniques, the computational hardware capabilities, as well as the gap between CFPD prediction capabilities and the needs of toxicologists and health-care providers. Specifically,


This section shows the roadmap to bridge the current simulation techniques to emerging toxicological/pharmaceutical paradigms to overcome the limitations of current numerical simulations. The long-term goal is to build an accurate and cost-effective digital respiratory simulation framework of subject-specific exposure to inhalable aerosols under realistic breathing cycles. A proposed framework and kinetic process of this multiscale model, i.e., a hybrid CFPD-LBM-PBPK model (see Sections 4.1–4.3), is shown in **Figure 11**. Details of future researches for extensions and improvements of the current CFPD model and the development of the next-generation model are discussed below.

observed that with the same initial diameter, the total deposition fraction for the three particles is EC-droplets #B2 < CSPs #1 < solid particles. Clearly, the hygroscopic growth of droplets and CSPs limited their Brownian diffusion, thereby reducing TDE in G3–G6. Since this effect is more prominent for EC-droplets than for CSPs (see **Figure 10**), the latter therefore has higher total deposition efficiency values. In other words, CSPs #1 deposit more in the G3–G6 TBU than #B2 EC-droplets due to lower hygroscopic growth and stronger Brownian diffusion effect. It can thus be conjectured that EC-droplets may deposit more than CSPs #1 up to the first bifurcation, due to more pronounced inertial impaction. In contrast, those EC-droplets tend to deposit more after the first bifurcation in G3–G6 TBUs and deeper lung regions where Brownian diffusion dominates deposition. Results also indicate the important effect of

hygroscopic behaviors of droplets on the accuracy of deposited-dose predictions.

Challenges faced by the current CFPD models and numerical studies for unconventional lung aerosol dynamics are mostly induced by the point-mass and point-force assumptions, the resolution of medical imaging techniques, the computational hardware capabilities, as well as the gap between CFPD prediction capabilities and the needs of toxicologists and health-care

**1.** The point-mass and point-force assumptions raise the difficulty to apply CFPD modeling techniques on aerosol transport and deposition simulation in the alveolar region, where the particle size will not be much smaller than a single mesh element and the assumptions

**2.** The current resolution of medical imaging techniques is not sufficiently high to identify small lung airways beyond G8. Therefore, most CFPD models are still for human upper airways, and it is impossible to reconstruct a complete subject-specific human respiratory system using CT/MRI imaging data. This restricts CFPD models from simulating aerosol transport in a full inspiratory-expiratory breathing cycle, i.e., the quantitative prediction

**3.** Due to the computational resources limitation, most CFPD models still use simplified mechanisms to reach good compromises between accuracy and computational efficiency, e.g., one-way coupling for dilute particle suspensions. Therefore, CFPD models are not suitable for dense particle suspensions in which the particle-particle interactions are not

**4.** Although CFPD models are able to provide high-resolution lung deposition patterns for unconventional aerosols, the deposited dose cannot be used directly for health risk assessment or drug delivery efficacy evaluation. Improvements need to be made for the

This section shows the roadmap to bridge the current simulation techniques to emerging toxicological/pharmaceutical paradigms to overcome the limitations of current numerical

current CFPD models to calculate the tissue and delivered doses.

**4. Challenges and future directions**

of exhalation clearance of aerosols.

providers. Specifically,

68 Aerosols - Science and Case Studies

are not valid.

negligible.

**Figure 11.** Schematic diagrams of the proposed multiscale CFPD-LBM-PBPK model: model framework with a complete human respiratory system geometry and kinetic processes after aerosol inhalation.

#### **4.1. Reconstructing and modeling the entire conducting zone of human lung**

In order to provide full-scale regional lung deposited doses in the oral/nasal cavity, conducting and respiratory zone, it is necessary to find methodologies to represent or reconstruct the complete human respiratory system geometry, including the smaller airways, which are not visible via CT/MRI imaging. Accordingly, two methods can be further developed: (1) coupling methods with single-path or multiple-path lung airway trees and (2) virtual lung model with idealized airways for higher generations covering the entire lung conducting zone (G0–G17). This is crucial to create a full-scale lung aerosol dynamics simulation featuring full breathing cycles.

#### *4.1.1. Coupling methods with single-path or multiple-path lung airway trees*

Coupling methods combine coupled boundary conditions with truncated lung airway trees to represent the whole conducting zone. Realization of coupling methods can be performed in the following steps:


Kleinstreuer and Zhang [40] used the TBU of Weibel type A and constructed the symmetric and in-plane tracheobronchial airways from G0 to G15, and used coupling method to simulate airflow and particle deposition in the lung airways with steady-state inhalation flow rate *Q*in = 30 L/min. Walters and Luke [42] created an 8-generation airway tree with random connection angles following Weibel's lung morphology. They applied coupled pressure boundary conditions for inhalation cases, and validated the coupling method for inhalation cases. They extended the application of this coupling method on particle transport in lung airways [43]. Tian et al. [41] developed the stochastic individual path (SIP) and multiple stochastic individual path (MSIP) approaches to represent the whole-lung geometry. Compared to the coupling method applied on detached airway geometries [41], this model is able to simulate transient breathing scenarios. Recently, similar models have also been developed [44, 45].

Since modeling the whole lung is computationally expensive, using the coupling method is potentially advantageous for fast-solution purposes. However, accuracy of the coupling methods still needs to be tested compared with using virtual lung model containing the entire conducting zone without any truncations.

#### *4.1.2. Virtual lung model for the entire conducting zone*

To encompass the entire conducting zone, the main objective of virtual lung model reconstruction is to develop a stochastic algorithm that generates 3D asymmetric human lower lung airways, which can then be connected to the subject-specific upper lung airways [46–48]. To establish such a virtual lung model, the upper airway (i.e., mouth and nasal airways, throat, and trachea) will be exactly reconstructed using image processing software that converts CT/ MRI images. Conducting airways, from the trachea down to the level of the terminal bronchioles, will be generated via a stochastic algorithm [46–48]. Specifically, by generating geometric variables (i.e., airway diameters, branching angles, and rotation angles of successive branching planes) stochastically with morphometric restrictions [46], following fractal model (i.e., Hess-Murray law), or grow from 1-D centerline airway tree [48], a complete tracheobronchial airway tree starting from the trachea can be constructed. The accuracy of the geometry can be judged in two ways: (1) the airways follow the centerlines of a 1D lung airway tree and (2) it conforms to surface data of the airways segmented from medical imaging. The combined geometry can be exported as a stereolithography (STL) file that can be used for mesh generation as needed for computer modeling and simulations.

Compared with coupling methods, constructing a virtual lung model will enable CFPD simulations with more realistic boundary conditions. It will be computationally achievable to simulate the multiphase inhaled aerosols in the entire lung conducting zone, with foreseeable increase in computational power.
