Proximal policy optimisation based black-box optimizer for tailoring inhalation instructions of an inhaler for minimising mouth-throat deposition.

Inhalation therapy using dry-powder inhalers (DPIs) offers targeted drug delivery for asthma and chronic obstructive pulmonary disease (COPD). Variation in deposition rate in the mouth-throat region depends on patient inhalation manoeuvres and techniques. The present study aimed to identify optimal combinations of head position, oral opening size, and mouthpiece insertion depth for three representative inhalation manoeuvres, corresponding to low, medium and high peak inspiratory flow rates (PIFRs). An idealised adult airway model was reconstructed and coupled with transient computational fluid dynamics (CFD) simulations that incorporated the Aerolizer® mouthpiece geometry, and inhaler resistance. A proximal policy optimisation (PPO) based black-box optimizer (BO) used deposition rate in mouth-throat (DEthroat) as the reward signal to guide optimisation toward minimisation of mouth-throat deposition in the Alberta Idealised Throat (AIT) model. Key findings indicate that optimised technique adjustments can markedly reduce mouth-throat deposition. For the low PIFR case, total DEthroat decreased from (28.82 ± 0.39) % in the original geometry to (5.11 ± 0.11) %. For the medium and high peak flow cases, average absolute reductions of 23.40 and 21.02 percentage points were achieved, respectively. Optimal actions across the three manoeuvres involved extending the head upward by 8.60°-13.21°, rotating left or right by 15.18°-21.9°, widening the oral opening to a three-finger width, and inserting the mouthpiece deeper by 7.99 mm-11.82 mm, based on the AIT model. These model-based findings suggest that head position and mouth-opening adjustments may influence aerosol transport and deposition patterns. The findings are hypothesis-generating and require validation in heterogeneous, subject-specific airway anatomies before clinical applicability can be inferred.