Machine Learning-Based Design of 3D-Printable Microneedles for Enhanced Tissue Anchoring with Reduced Tissue Damage

Abstract

Microneedles offer a minimally invasive and painless approach to drug delivery, enhancing therapeutic efficacy while minimizing side effects. Despite extensive efforts to minimize tissue damage and pain during microneedle insertion, identifying an optimal shape that simultaneously enhances tissue anchoring and reduces insertion force remains challenging, as anchoring typically relies on barbed features that inherently increase tissue damage. This study presents a machine learning-based methodology for optimizing microneedle shapes, combining Bayesian optimization and finite element methods to develop novel, low-pain, low-damage, and improved tissue-anchoring microneedle designs. Specifically, the study focuses on optimizing designs that are readily manufacturable via 3D printing to ensure scalable and cost-efficient production. Compared to conventional cone-shaped microneedles, the optimal design increases the pull-out-to-penetration energy ratio by 5.2-fold, reflecting enhanced tissue anchoring and reduced tissue damage. For validation, we conducted experiments and confirmed a 6.0-fold improvement in PPR in the optimized microneedle design. By incorporating application-specific modifications, this approach has the potential to generate diverse, minimal pain microneedle designs tailored to a variety of medical applications.