Machine Learning-Enabled Optimization of Interstitial Fluid Collection via a Sweeping Microneedle Design

Microneedles (MNs)allow for biological fluid sampling and drugdelivery toward the development of minimally invasive diagnosticsand treatment in medicine. MNs have been fabricated based on empiricaldata such as mechanical testing, and their physical parameters havebeen optimized through the trial-and-error method. While these methodsshowed adequate results, the performance of MNs can be enhanced byanalyzing a large data set of parameters and their respective performanceusing artificial intelligence. In this study, finite element methods(FEMs) and machine learning (ML) models were integrated to determinethe optimal physical parameters for a MN design in order to maximizethe amount of collected fluid. The fluid behavior in a MN patch issimulated with several different physical and geometrical parametersusing FEM, and the resulting data set is used as the input for MLalgorithms including multiple linear regression, random forest regression,support vector regression, and neural networks. Decision tree regression(DTR) yielded the best prediction of optimal parameters. ML modelingmethods can be utilized to optimize the geometrical design parametersof MNs in wearable devices for application in point-of-care diagnosticsand targeted drug delivery.