Machine learning and sensitivity analysis for predicting nasal drug delivery for targeted deposition

Area of Science:

• Pharmacokinetics and Drug Delivery
• Computational Fluid Dynamics
• Biomedical Engineering

Background:

• Targeted nasal drug delivery offers enhanced efficacy but is influenced by administration technique and device parameters.
• Optimizing nasal drug delivery requires understanding complex interactions between numerous variables.
• Computational modeling is essential for exploring the vast parameter space influencing particle deposition.

Purpose of the Study:

• To investigate the impact of various spray parameters on drug particle deposition in different nasal regions.
• To evaluate the effectiveness of machine learning models in predicting nasal drug deposition.
• To identify key parameters significantly affecting drug delivery efficiency in the nasal cavity.

Main Methods:

• Simulated 384 unique spray characteristic combinations by varying six input parameters (e.g., spray angle, velocity, particle size) and three inhalation flow rates.
• Utilized a time-averaged frozen flow field approach with Large Eddy Simulation to reduce computational cost.
• Tracked particle trajectories to determine deposition in anterior, middle, olfactory, and posterior nasal regions.

Main Results:

• Particle size distribution was a significant factor for deposition in olfactory and posterior nasal regions.
• Spray device insertion angle critically influenced deposition in the anterior and middle nasal regions.
• Machine learning models demonstrated accurate predictive capabilities for particle deposition, even with a limited dataset.

Conclusions:

• Spray characteristics, particularly particle size and insertion angle, play crucial roles in nasal drug deposition patterns.
• Computational simulations coupled with machine learning provide a viable approach for optimizing nasal drug delivery systems.
• Further research can leverage these findings to design more effective nasal drug formulations and delivery devices.