Artificial intelligence for stromal cell quality assessment: A systematic review of human mesenchymal stromal cell properties and viability.

Reliable assessment of mesenchymal stromal cell (MSC) viability and stem cell-associated properties is essential for ensuring the consistency, safety, and therapeutic potential of MSC cultures before clinical use. However, incomplete understanding of the mechanisms underlying MSC therapeutic effects remains a major barrier to clinical translation and to establishing reliable, standardised quality attributes. Artificial intelligence (AI) may enable automated assessment of these properties, but the performance and clinical relevance of AI-based approaches have not been systematically evaluated. This systematic review evaluated the accuracy and clinical utility of AI for assessing viability and stem cell-associated properties in human MSC cultures. Following PRISMA guidelines, we searched five electronic databases (Cochrane, Scopus, PubMed, EMBASE, and MEDLINE) for eligible studies reporting quantitative AI-based assessments of human MSC cultures. Two reviewers independently screened studies, extracted data, and assessed risk of bias using PROBAST. The review protocol was prospectively registered with PROSPERO (CRD420256393235). Of 3,872 records identified, 18 studies met the inclusion criteria. Supervised AI models, particularly convolutional neural networks, were most used. Across included studies, reported performance metrics varied substantially because of differences in AI models, imaging modalities, datasets, and outcome definitions. Considerable heterogeneity in AI model architectures, imaging modalities, outcome definitions, and reporting metrics limited direct comparison across studies. Overall, available evidence suggests that AI can assess key MSC quality attributes and may support more objective, scalable, and standardised quality-control workflows. However, broader validation, standardised reporting, external testing, and evaluation across a wider range of MSC characteristics are required before routine clinical implementation can be recommended.