Year: 2025
Type: article
Abstract: Stem cell-based therapies have potential in the treatment of many degenerative and auto-immune diseases, yet the success of such therapies has been hampered by the heterogeneity of cultured cell populations. Traditional sorting, including fluorescence-activated sorting (FACS) and magnetic-activated sorting (MACS) are based on labeling technologies that may induce cell damage, introduce variance, or retain residual reagents into a therapeutic product. Here we introduce a label-free cell sorting system using artificial intelligence (AI) that combines microfluidic imaging and decision making based on deep learning to increase the accuracy and safety of stem cell therapy. The system uses real-time, high-throughput cell-in-flow imaging, and lightweight convolutional neural networks to classify subpopulations on the basis of morphology, texture, and deformability cues. Millisecond-scale actuation of the selective isolation of therapeutically potent subsets can be performed with intelligent automation, including mesenchymal stem cells of high immunomodulatory potential or induced pluripotent stem cell derivatives of low tumorigenic risk. The AI-assisted sorter enhances purity, viability and functional consistency compared to traditional methods, and reduces batch to batch variation. Such strategy offers a translational route to standardized, scalable, and safe stem cell commodities, with smart automation being a key facilitator of future-generation regenerative drug.
DOI: 10.64206/14rf5w80
Author: Adedoyin Adetoun Samuel
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4413873525