Authors
Md Saimun Alam, Somaiyeh Khoubafarin Doust, Aniruddha Ray
Published in
Journal of biophotonics. Volume 19. Issue 11.
Abstract
Cytotoxicity assessment is fundamental to many areas of biomedical research, yet cell viability assays alone do not reveal death mechanisms, for example, apoptosis and necrosis. Conventional endpoint assays, including Western blotting, ELISA, and fluorescence microscopy/cytometry, require costly reagents, extensive sample preparation, and specialized expertise, limiting accessibility and scalability. Here, we present a multimodal spatiotemporal deep learning framework for classifying cell death mechanisms from phase-contrast and dark-field microscopy. Time-lapse videos of drug-treated BT-20 breast cancer cells were recorded in both modalities. These cells were used to train 3D convolutional, convolutional-recurrent, transfer-learning, and transformer models. Combined across architectures and modalities, the models achieved a balanced accuracy of 0.98 on recordings not used for training. The CytoMorpheus Analyzer runs the full pipeline from raw microscopy data, classifying over a thousand cells within minutes. These results highlight the potential of label-free microscopy with deep learning for real-time, high-throughput profiling of cell death and drug response.
PMID:
43049560
Bibliographic data and abstract were imported from PubMed on 18 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 6
- Comments 0