| dc.description.abstract | Chronic disorders are leading causes of disability and death, and their high pervasiveness calls for effective treatments leading to a long-term cure. Studies show that stem cell-based therapies can provide successful solutions to chronic diseases. Mesenchymal stromal cells (MSCs), a multipotent group of stem cells, are extensively used for cell-based therapies owing to their immunomodulation, ex vivo proliferation, and clinically significant effects. MSC-based solutions require manufacturing large volumes of viable cells that is highly dependent on reliable methods to characterize cell mechanisms. Previous findings have revealed that cell morphology can serve as a critical quality attribute to predict the therapeutic potency of MSCs. Current standards to estimate MSC effectiveness based on their morphological phenotype are subjective, destructive, or time-consuming. Therefore, an objective method for morphological screening is needed to analyze the viability of live MSC cultures non-invasively. Computer-aided image analysis is an excellent tool to extract relevant cellular features rapidly and efficiently. This dissertation aims to facilitate large-scale cell growth strategies by providing automated technologies for evaluating images of cells grown in monolayer and three-dimensional environments.
We developed an analysis scheme based on conventional image processing techniques combined with machine learning to examine monolayer cultured MSCs. This methodology proved the applicability of image analysis as a robust tool for quantifiable culture monitoring. To address the challenges of standard image analysis approaches, we explored deep learning to better identify individual cells. It showed improved results for cell localization and also optimized overall MSC assessment. Lastly, we extrapolated deep learning-based analysis to study three-dimensional cultures used to produce MSCs in commercially viable quantities. This work shows great promise to fulfill the unmet need for cytomorphological analysis and downstream representation of MSCs adhered to spherical microcarriers. The algorithms were validated using visual inspection by a bi-ologist with 15+ years of experience with MSCs. Thus, our research exhibits valuable potential to quantitatively analyze MSC efficacy and functionality, enabling the advancement of cytotherapies. | |
