Related Experiment Video
Updated: May 2, 2026

12:55
Strategies for Tracking Anastasis, A Cell Survival Phenomenon that Reverses Apoptosis
Published on: February 16, 2015
17.9K
Apoptosis detection for non-adherent cells in time-lapse phase contrast microscopy
1Lane Center for Computational Biology and Robotics Institute, Carnegie Mellon University, USA. seungilh@cs.cmu.edu
Summary
This study introduces a vision-based method to detect apoptosis, or programmed cell death, in non-adherent cells. The approach accurately tracks cell transitions from live to dead, aiding in monitoring cell expansion and hematopoietic stem cell therapies.
Area of Science:
- Biotechnology
- Cell Biology
- Medical Imaging
Background:
- Apoptosis (programmed cell death) is crucial for non-perturbative cell expansion monitoring.
- Detecting apoptosis in non-adherent cells presents unique challenges compared to adherent cells.
- Hematopoietic stem cells (HSCs) are vital for bone marrow transplants.
Purpose of the Study:
- To develop a vision-based method for detecting apoptosis in non-adherent cells.
- To accurately determine the occurrence and timing of cell death events.
- To validate the method's performance using hematopoietic stem cell populations.
Main Methods:
- Cell regions are detected and tracked over time to create cell tracklets.
- Visual properties of each tracklet are analyzed to identify live-to-dead cell transitions.
- A transductive learning framework is employed, utilizing both labeled and unlabeled data.
Main Results:
- The vision-based method successfully detects apoptosis in non-adherent cells.
- The approach accurately determines the timing of cell death events.
- Promising performance was observed in experiments with hematopoietic stem cell populations.
Conclusions:
- The proposed vision-based method offers a non-perturbative approach for monitoring apoptosis in non-adherent cells.
- This technique has potential applications in cell therapy, particularly with hematopoietic stem cells.
- The use of transductive learning enhances the method's robustness by incorporating unlabeled data.

