3D Assessment of Immunomodulatory Effects of Cytokine-treated Mesenchymal Stromal Cells Using AI-Based Organ and Tissue Detection Methods 2309067

Abstract Introduction Cell tracking and biodistribution studies are constrained by slow, labor-intensive workflows involving manual tissue segmentation and limited 3D visualization. BioInVision’s CryoViz™ creates high-resolution, microscopic color anatomical and molecular fluorescence volumes enabling 3D visualization and quantification of fluorescently labeled cells across a whole animal. Tissue Recovery and Molecular Analysis (TissueReM), an AI model for tissue segmentation, delivers repeatable biodistribution readouts that are faster than conventional harvesting-based assays. We applied our workflow consisting of CryoViz™ fee-for-service imaging and TissueReM AI-based segmentation to obtain 3D biodistribution analysis of cytokine treated Umbilical Cord Mesenchymal Stromal cells (ucMSCs) in whole mice. ucMSCs have immunomodulatory properties i.e. they modify T cell and NK cell behavior and play an important role in immunotherapy applications. Methods We intravenously injected mice with fluorescently labeled ucMSCs and imaged them with CryoViz™. From these image volumes, we segmented tissues by applying TissueReM’s convolutional neural network (CNN) AI model trained on paired CryoVizTM brightfield images and human expert segmented masks. We chose Dice Score, an accuracy metric that measures overlap with human expert ground truth, and organ-specific cell counts derived from AI to those obtained from human expert generated masks, to determine model performance. Results This resulted in Dice scores of 0.80 (lung), 0.91 (liver) and 0.83 (spleen) - major tissues analyzed in this study. TissueReM-derived cell counts were within 10% of the counts derived from human expert masks. This workflow preserves whole-body spatial context while reducing analysis time and improving consistency. Conclusion CryoViz™ Imaging combined with TissueReM is a highly efficient scalable alternative to traditional biodistribution analysis in cell therapy, supporting rapid optimization of dose and route while retaining anatomical context. Funding Source NIH/NIGMS Topic Categories Technological Innovations in Immunology (TECH)