The use of dry powder inhalers (DPIs) represents a cornerstone in the treatment of chronic pulmonary diseases. However, suboptimal inhalation techniques, including inadequate airflow rates, have been a persistent concern for achieving effective therapeutic outcomes, as many patients remain unaware of their insufficient inhalation performance. As an effective strategy, a digital monitoring system, coupled with dry powder inhalers (DPIs), has emerged to estimate flow profiles and provide inhalation information. The estimation could be further facilitated by the application of artificial intelligence (AI). In this work, a novel digital system to primarily monitor pressure during DPI usage was successfully designed, and advanced machine learning (ML) techniques were then employed to estimate inhalation flow profiles based on the captured data. Four optimal machine learning models were selected for subsequent inhalation parameter prediction, given their superior generalization ability. By using these models, inhalation flow profiles could be successfully estimated, with an excellent accuracy of 97.7% for Peak Inspiratory Flow Rate (PIFR) and 95.2% for inspiratory capacity (IC). In summary, the pressure-based digital monitoring system empowered by AI techniques could be successfully applied to assess inhalation flow profiles with excellent accuracy.
Dry powder inhalers (DPIs) are essential for pulmonary drug delivery, yet how adhesive formulation powders are converted into well‐dispersed aerosols remains poorly resolved, which is highly device‐ and formulation‐dependent. Here we integrate device and formulation characterization, theoretical analysis, numerical simulations, and machine learning to map airflow–particle dynamics across DPIs. Device geometry sets internal flow fields, which modulate drag, swirl, and impact, thereby mobilizing powders and driving transient deagglomeration. For fine particle‐dominated DPI formulations, gravity acting on fines is negligible, whereas adhesive force dominates in static powder beds and decays as agglomerates break up, driven by aerodynamic interactions and particle impacts. Highly adhesive carrier‐free agglomerates disperse via strong swirl and frequent particle–wall impacts, while carrier‐based blends with moderate adhesion rely on aerodynamic drag and high‐energy carrier impacts. Further assisted by machine learning, we identify flow rate, drag, and particle–wall impacts as primary contributors to dispersion, whereas particle–particle impacts are a principal inhibitor. This integrated work reveals how devices and formulations shape airflow–particle dynamics and aerodynamic performance, providing mechanism‐based guidance for next‐generation DPI design.