Machine Learning Driven Optimization for High Precision Cellular Droplet Bioprinting

Jaemyung Shin,Minseok Kang,Kinam Hyun,Zhangkang Li,Hitendra Kumar,Kangsoo Kim,Simon S. Park,Keekyoung Kim

DOI: https://doi.org/10.1101/2024.09.04.611131

2024-09-08

Abstract:Controlled volume microliter cell-laden droplet bioprinting is important for precise biologics deposition, reliably replicating 3D microtissue environments for building cell aggregates or organoids. To achieve this, we propose an innovative machine-learning approach to predict cell-laden droplet volumes according to input parameters. We developed a novel bioprinting platform capable of collecting high-throughput droplet images and generating an extensive dataset for training machine learning and deep learning algorithms. Our research compared the performance of three machine learning and two deep learning algorithms that predict droplet volume based on numerous bioprinting parameters. By adjusting bioink viscosity, nozzle size, printing time, printing pressure, and cell concentration as input parameters, we precisely could control droplet sizes, ranging from 0.1 µL to 50 µL in volume. We utilized a hydrogel precursor composed of 5% gelatin methacrylate and a mixture of 0.5% and 1% alginate, respectively. Additionally, we optimized the cell bioprinting process using green fluorescent protein-tagged 3T3 fibroblast cells. These models demonstrated superior predictive accuracy and revealed the interrelationships among parameters while taking minimal time for training and testing. This method promises to advance the mass production of organoids and microtissues with precise volume control for various biomedical applications.

Bioengineering

What problem does this paper attempt to address?

The problem this paper attempts to address is how to precisely control droplet volume in cell-laden droplet bioprinting to reliably replicate 3D microtissue environments for constructing cell aggregates or organoids. Specifically, the authors propose an innovative machine learning approach to predict the volume of cell-laden droplets based on input parameters and develop a new bioprinting platform capable of collecting high-throughput droplet images, generating large datasets for training machine learning and deep learning algorithms. By adjusting parameters such as bioink viscosity, nozzle size, printing time, printing pressure, and cell concentration, the authors achieved precise control over droplet size, ranging from 0.1 µL to 50 µL. Additionally, the authors optimized the bioprinting process using 3T3 fibroblasts labeled with green fluorescent protein.

These models demonstrated excellent predictive accuracy and revealed the interrelationships between parameters while requiring minimal time during training and testing. This approach is expected to drive the large-scale production of organoids and microtissues, providing precise volume control for various biomedical applications.