YOlOv4 ML Image Detection for Melanoma Cancer Cell Counting Presenter
Session Number
CMPS(ai) 20
Advisor(s)
Sonbinh T. Nguyen and Nakisha Shanel-Latanya Rutledge,Northwestern University
Discipline
Computer Science
Start Date
17-4-2025 11:25 AM
End Date
17-4-2025 11:40 AM
Abstract
Machine learning(ML) models are computer systems that are able to learn and adapt independently, using algorithms to analyze and draw inferences from patterns in data. Usage of ML models in biology has skyrocketed, especially in the drug discovery and pharmaceutical industries. Cell counting is an important part of the drug discovery process, crucial for assessing culture viability and determining proliferation rates post treatment. Human cell counters are prone to bias and error, leading to wasted time for accurate results. Additionally, drug compounds often interfere with commercially available biochemical cell assays used for assessing the above-mentioned metrics. In this project, we use the YOLOv4 machine learning model to automate cell counting via image object detection. The model was trained on images of mouse melanoma cell line B16F10, human uveal melanoma, 624.38, and 888.A2 melanoma cell lines post nanoparticle treatment, captured by the invitrogen M5000 Imaging System and Incucyte Imaging Software. Prior to training, the images were converted to jpg format and then cells were labeled either dead or alive using the labelImg python program. Using this model, we can count cells more effectively, resulting in a precise, cost-effective, time-saving, and bias-free alternative to manual counting.
YOlOv4 ML Image Detection for Melanoma Cancer Cell Counting Presenter
Machine learning(ML) models are computer systems that are able to learn and adapt independently, using algorithms to analyze and draw inferences from patterns in data. Usage of ML models in biology has skyrocketed, especially in the drug discovery and pharmaceutical industries. Cell counting is an important part of the drug discovery process, crucial for assessing culture viability and determining proliferation rates post treatment. Human cell counters are prone to bias and error, leading to wasted time for accurate results. Additionally, drug compounds often interfere with commercially available biochemical cell assays used for assessing the above-mentioned metrics. In this project, we use the YOLOv4 machine learning model to automate cell counting via image object detection. The model was trained on images of mouse melanoma cell line B16F10, human uveal melanoma, 624.38, and 888.A2 melanoma cell lines post nanoparticle treatment, captured by the invitrogen M5000 Imaging System and Incucyte Imaging Software. Prior to training, the images were converted to jpg format and then cells were labeled either dead or alive using the labelImg python program. Using this model, we can count cells more effectively, resulting in a precise, cost-effective, time-saving, and bias-free alternative to manual counting.