Files
Abstract
The increasing prominence of cell therapy in oncological treatment necessitates innovative, non-destructive methods for efficient cell manufacturing. Current evaluation techniques are often invasive or incompatible with cell therapies, a limitation addressed by Quantitative Phase Imaging methods such as Differential Phase Contrast (DPC). DPC enables high-resolution, label-free imaging of live cells. In combination with machine learning algorithms, this technology permits real-time, non-invasive evaluation of therapeutic cells, including Mesenchymal Stem Cells (MSCs) and Chimeric Antigen Receptor T-cells (CAR-T). A study was conducted involving anti GD2 and mCherry CAR-T cells, employing DPC and Fourier Ptychography for imaging over three days. Images were reconstructed using Python, and cellular segmentation was conducted via CellProfiler. Subsequent machine learning analysis, particularly through the use of KNN, Random Forest and an autoencoder, the latter of which achieved classification accuracies of up to 93% on aggregate data, thereby validating this integrated approach for quality control in therapeutic cell manufacturing.