Document Type
Report
Abstract
Important cell processes can be captured and analyzed with live-cell imaging followed by computational analysis. Current methods of microscopy image analysis involve manually segmenting cells, often done by multiple researchers, which is time-consuming and can limit the accuracy and throughput of the desired data. By implementing deep learning algorithms, captured images can be automatically segmented which greatly reduces the time required to do so and can possibly improve the accuracy of desired data. A newly developed deep learning algorithm pipeline, Lipid Locator, will be utilized to automatically segment acquired images of adipose-derived stem cells (ADSC). More specifically, ADSC with fluorescently labeled nuclei and lipid droplets during adipogenic differentiation (AD). Benchmark data testing will be done to compare the pretrained models of Lipid Locator’s automated cell segmentation data to manually segmented cells of the same images. When comparing automated cell segmentation to manual segmentation, the morphological characteristics of ADSC will also be analyzed such as cell count, nuclear area, and lipid accumulation area during AD. This provides an opportunity to quantify automated and manual cell segmentation benchmark data.
Publication Date
5-2024
Faculty Mentor of Honors Project
Dr. Michael Cho
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Armstrong, Zachary, "Stem Cell Potency: Automated vs. Manual Segmentation in Adipogenic Differentiation" (2024). 2024 Spring Honors Capstone Projects. 3.
https://mavmatrix.uta.edu/honors_spring2024/3