Poster 30 - Automating Reactive Astrogliosis Quantification: A Python-Based Approach for Cellular Analysis

Saturday, October 26, 2024

10:00 AM - 12:00 PM EST

Location: Exhibit Hall E, F & G

Presenting Author(s)

CL

Connor Lee

Volunteer Research Intern
Institute of Engineering in Medicine, United States

Co-Author(s)

AL

Albert Li

volunteer research intern
Outreach Program for Advancing Learning in STEM (OPALS), United States
HH

Huiying Huang, n/a

Undergraduate student
University of California, San Diego, United States
Yidam Park

volunteer research intern
Outreach Program for Advancing Learning in STEM (OPALS)
San Jose, California, United States
William Chang

High School Research Intern
Outreach Program for Advancing Learning in STEM (OPALS)
Alhambra, California, United States
MZ

Mimi Zhao

volunteer research intern
Outreach Program for Advancing Learning in STEM (OPALS), United States
RY

Ryan Yang

volunteer research intern
Outreach Program for Advancing Learning in STEM (OPALS), United States
Shawn Kim (she/her/hers)

volunteer research intern
Outreach Program for Advancing Learning in STEM (OPALS)
San Diego, California, United States

Last Author(s)

VG

Veronica Gomez-Godinez

Scientist
University of California, San Diego
La Jolla, California, United States

Primary Investigator(s)

Linda Shi, PhD (she/her/hers)

OPALS Program Director, Project Scientist, Co-director IEM
University of California, San Diego: Institute of Engineering in Medicine
La Jolla, California, United States

Introduction: Reactive astrogliosis is a critical process in the central nervous system's response to injury, aiding to repair neural cells after incidences. This phenomenon can be studied in the context of controlled injuries, like those created by laser-induced shockwaves (LIS), which mimics blast-induced traumatic brain injury. Quantifying the number of cells undergoing reactive astrogliosis following such injuries is essential for understanding neural repair mechanisms. Traditional methods of quantifying these cells often involve labor-intensive manual counting, which can be prone to errors and is very time-consuming. To address these challenges, this project focuses on the development of a Python script designed to automate the quantification of reactive astrocytes. Using artificial intelligence and machine learning, the model was trained on various fields of views (FOV) from different datasets, and the code aims to accurately detect and count astrocytes, streamlining the analysis process and reducing human bias. This approach not only enhances the efficiency of data analysis but also provides a solution for handling large datasets in research.

Materials and

Methods: Mouse glial cultures isolated from the cortex were subjected to laser-induced shockwave (LIS) after 93 days in vitro and fixed with 4% paraformaldehyde to stain against GFAP-1 and IBA-1. After the cells were shockwaved, we captured images of each FOV, and initially conducted an analysis of each position individually to determine the cell count. However, to address the inefficiencies of manually counting each field of view, we developed a Python script to accelerate our data analysis.

Results, Conclusions, and Discussions: The Python script is also able to take over the task of stitching FOVs together, significantly reducing the time and hardware requirements to perform this task in FIJI/ImageJ. Additionally, the user is able to click on a desired FOV to automatically perform analysis, which takes only 10 seconds to run compared to several minutes if done manually. The code is able to achieve this with only a 8% reduction in accuracy while speeding up the process by many times, significantly increasing competence compared to our previous methods, which took about 12 hours per data set when done manually. Future work should focus on enhancing the efficiency and accuracy of the code, as it currently requires 1.5 hours to process a complete data set with an accuracy of 92%. Another item we could improve on is the current model that is being used to identify cells. Expanding the training data would allow for much higher accuracy.

Acknowledgements (Optional): This work was supported by a generous gift from Beckman Laser Institute Inc. to L.S. and V.G.G. The high school students (*) participated in the UCSD IEM OPALS internship. Special thanks to Dr. Shu Chien from UCSD Bioengineering and Dr. Lizhu Chen from CorDx Inc. for their donations.

References (Optional):