Honors Theses
Advisor
Russell Pirlo, Ph.D.
Department
Chemical and Materials Engineering
Publication Date
4-1-2024
Document Type
Honors Thesis
Abstract
Many tools have emerged to investigate the functioning of biological systems, especially when in contact with foreign substances. In vitro procedures are often used due to their cost effectiveness and suitability for high-throughput experiments. These procedures collect basic measurements, such as toxicity and biocompatibility, that provide insight into the compatibility and safety of a substance. In vitro toxicity tests are favored for their expediency, affordability, and consistent outcomes. Quantitative methodologies, like colorimetric and fluorometric assays, offer objectivity and high-throughput analysis. However, they require lengthy incubation times and only provide a single metric. Microscopy-based methods provide more information in terms of cell morphology and localization and can be captured quickly without the need for reagents and incubation. Yet, this method requires specialized expertise and is prone to subjective biases and variations based on the region of interest. Given the limitations of microscopy-based approaches, there is a growing interest in leveraging machine learning (ML) to streamline and enhance cell analysis. This study aims to develop an ML-based approach to evaluate cell count and confluency from microscope images and compare its performance to the colorimetric assay, CCK8. The CCK8 assay, which releases a dye when metabolized by live cells, served as the benchmark for comparison. The ML-based method developed using Ilastik, CellProfiler, and Python, segments microscopy images into cell and background regions, followed by erosion for cell boundary enhancement. CellProfiler subsequently quantifies the cell count and confluency from the processed images. This novel ML-based approach offers expedited analysis, while mitigating the inherent subjectivity and error associated with conventional techniques. This approach also eliminates the need for excess reagents and waste associated with quantitative assays. In conclusion, this technique presents an alternative in scenarios where traditional assays are impractical, such as with low cell counts or when cells must be reused.
Permission Statement
This item is protected by copyright law (Title 17, U.S. Code) and may only be used for noncommercial, educational, and scholarly purposes.
Keywords
Undergraduate research
eCommons Citation
Jones, Adam J., "Development of a Machine Learning-Based Program to Measure Cell Proliferation" (2024). Honors Theses. 446.
https://ecommons.udayton.edu/uhp_theses/446
COinS
