Related Experiment Video
Updated: Jan 11, 2026
06:03
AMEBaS: Automatic Midline Extraction and Background Subtraction of Ratiometric Fluorescence Time-Lapses of Polarized Single Cells
Published on: June 23, 2023
802
CellApop: A knowledge-guided decoupled distillation framework for label-efficient apoptotic cell segmentation and
Chi Dong1, Xuan Xie2, Shuai Yue2
1School of Intelligent Medicine, China Medical University, Liaoning 110122, PR China.
Computer Methods and Programs in Biomedicine
|November 12, 2025
Summary
This study introduces CellApop, a deep learning framework for label-free apoptosis detection in microscopy. It accurately quantifies apoptotic cells, outperforming junior experts and reducing manual labeling needs.
Area of Science:
- Biotechnology
- Cell Biology
- Artificial Intelligence
Background:
- Conventional apoptosis detection relies on labor-intensive, potentially cytotoxic fluorescence staining.
- Existing methods are unsuitable for real-time monitoring of cellular processes.
Purpose of the Study:
- To develop a label-free, dynamic detection framework for apoptotic cells using bright-field microscopy.
- To create a segmentation-based deep learning (DL) model for efficient and accurate apoptosis identification.
Main Methods:
- A Knowledge-guided Decoupled Distillation (KDD) framework trained a lightweight DL network (CellApop) on 16,472 bright-field cell images.
- CellApop incorporated advanced modules for improved segmentation accuracy in challenging conditions.
- Performance was validated using Dice similarity coefficient, Hausdorff Distance, and Intersection over Union.
Main Results:
- CellApop achieved high Dice scores (0.843 general, 0.754 apoptotic cells) with reduced complexity and latency.
- The KDD strategy reduced manual labeling by ~80% on a proprietary dataset.
- Automated apoptosis rates showed high concordance with ground truth and expert assessments.
Conclusions:
- CellApop provides accurate, efficient, label-free segmentation of apoptotic cells, negating the need for fluorescence staining.
- The framework is robust and scalable for automated apoptosis quantification and drug-response assessment.
- This DL approach offers a promising tool for routine experimental workflows in cell biology.
