Related Experiment Video
Updated: Apr 20, 2026
09:31
High-Speed Ultraviolet Photoacoustic Microscopy for Histological Imaging with Virtual-Staining assisted by Deep Learning
Published on: April 28, 2022
3.6K
Deeptosis: A Deep Learning-Based Platform for Label-Free Discrimination of Apoptosis and Pyroptosis from Brightfield
Haiji Wang1, Shaofeng Lin2, Chenbei Li3
1School of Biomedical Sciences, Hunan University, Changsha, China.
Journal of Molecular Biology
|April 18, 2026
Summary
Deeptosis, a deep learning pipeline, accurately distinguishes apoptosis and pyroptosis using bright-field images. This automated system aids in studying cell death mechanisms and their disease implications.
Area of Science:
- Cell biology
- Computational biology
- Immunology
Background:
- Distinguishing apoptosis from pyroptosis is crucial for understanding regulated cell death.
- Similar cell morphologies and shared signals complicate label-free bright-field discrimination.
Purpose of the Study:
- To develop an automated, label-free method for distinguishing apoptosis and pyroptosis.
- To provide a quantitative analysis framework for cell death modalities.
Main Methods:
- An end-to-end deep learning pipeline named Deeptosis.
- Utilizes Cellpose for automatic cell segmentation and a Vision Transformer (ViT) for classification.
- Trained on 26,565 manually annotated bright-field cells.
Main Results:
- Achieved a mean AUROC of 0.999 in cross-validation.
- High performance on an independent test set (AUROC 0.990 for apoptosis, 0.982 for pyroptosis).
- Outputs color-coded visualizations and per-cell CSV data.
Conclusions:
- Deeptosis offers a robust, label-free framework for quantitative cell death analysis.
- The system is accessible via a web interface for batch processing.
- Facilitates research in immunity and disease by enabling accurate cell death discrimination.
