Ensemble Machine Learning Outperforms Traditional Models for High-Accuracy Cell Viability Classification in Flow Cytometry Data

Flow cytometry is a cornerstone of modern biological research, yet its widespread adoption is hindered by cost and time constraints. This study explores machine learning (ML) to streamline flow cytometry workflows, focusing on classifying cell viability (live/apoptotic) using only morphological parameters (FSC/SSC). By evaluating ensemble models (Random Forest, XGBoost, CatBoost, LightGBM) and neural networks, it is demonstrated that ML achieves robust classification (ROC-AUC: ۰.۹۷, accuracy: ۹۱.۶%) while reducing reliance on fluorescence-based markers. The results highlight ML’s potential to enhance accessibility and cost-efficiency in flow cytometry.

Keywords:

flow cytometry , machine learning , cell viability , ensemble methods , computational efficiency