A framework to optimize spring-driven autoinjectors.

Area of Science:

• Biomedical Engineering
• Computational Modeling
• Drug Delivery Systems

Background:

• Autoinjector design faces challenges in accurate modeling and meeting diverse requirements.
• Optimizing devices for multiple drugs with varying viscosities is complex.
• Existing computational models can be computationally intensive.

Purpose of the Study:

• To develop a computational framework for optimizing spring-driven autoinjectors for multi-drug delivery.
• To enhance computational efficiency through surrogate modeling.
• To identify robust and optimal autoinjector designs.

Main Methods:

• Development of a computational model for spring-driven autoinjectors.
• Replacement of the computational model with a deep neural network surrogate model.
• Application of Sobol sensitivity analysis and multi-objective optimization using Pymoo.
• Evaluation of solution robustness.

Main Results:

• A deep neural network surrogate model achieved a 1,000-fold improvement in computational efficiency.
• Sensitivity analysis identified key input effects on autoinjector performance.
• Multi-objective optimization yielded Pareto optimal designs for multi-drug delivery.
• Two promising autoinjector designs were identified after robustness evaluation.

Conclusions:

• The proposed framework enables efficient and effective design optimization of autoinjectors for multiple drugs.
• The computational efficiency allows for easier incorporation of decision-maker preferences.
• This approach facilitates the development of advanced drug delivery devices.