Related Experiment Video
Updated: Jun 7, 2025
09:53
Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020
7.2K
Machine Learning Identification and Classification of Mitosis and Migration of Cancer Cells in a Lab-on-CMOS
IEEE Journal of Biomedical and Health Informatics
|November 12, 2024
Summary
This study introduces a computational framework for label-free cell culture assays using CMOS capacitance sensors. The framework accurately identifies cell mitosis and migration from capacitance data, enhancing single-cell analysis capabilities.
Area of Science:
- Biotechnology
- Cell Biology
- Sensor Technology
Background:
- Conventional cell assays (immunohistochemistry, immunofluorescence, flow cytometry) require labeling, limiting their use in integrated devices.
- Label-free cell culture assays using CMOS capacitance sensors offer a promising alternative for compact analytical systems.
- Existing capacitance sensor techniques can be enhanced for more detailed cellular behavior analysis.
Purpose of the Study:
- To develop a computational framework for augmenting CMOS capacitance sensors in cell culture assays.
- To enable label-free identification and classification of specific cell behaviors like mitosis and migration.
- To improve the analytical capabilities of capacitance sensors for single-cell analysis.
Main Methods:
- Utilized CMOS capacitance sensors for label-free cell culture monitoring.
- Engineered two novel time series features from capacitance data for cell behavior discrimination.
- Developed a computational framework to analyze capacitance time series data for mitosis and migration.
- Validated the feature representation technique across multiple experiments using leave-one-run-out testing.
Main Results:
- Achieved an Area Under the Curve (AUC) of 0.719 for cell behavior classification using the receiver operating characteristic (ROC) curve.
- Demonstrated the robustness and applicability of the feature representation technique across different experimental conditions.
- Obtained an F-1 score of 0.803 and a G-Mean of 0.647 in leave-one-run-out validation, confirming reliable classification.
Conclusions:
- The developed computational framework significantly enhances label-free cell culture assays using CMOS capacitance sensors.
- The engineered time series features effectively discriminate single-cell behaviors, including mitosis and migration.
- This approach offers a powerful tool for advanced cell analysis in compact, integrated systems without the need for labeling.
