The artificial intelligence to analyze and characterize cell lines based on the processes of visible spectrum image data.

Abstract

Abstract In particular, as part of the extensive Stem Line project Mito-Cell-UAB073 «Stem cell lines - Quality control», we have taken a specific interest in a new and complementary CQ approach to cell line and stem cell line intelligence in Biobank. We have combined computer vision image processing methods and deep learning techniques to create the non-invasive Life Cell AI UAB model for robust prediction of cell line viability, using single static images obtained from standard optical light microscope systems. The Life Cell AI UAB model showed a sensitivity of 82.1% for viable cell lines while maintaining a specificity of 67.5% for non-viable cell lines across three independent blind test sets from different biotechnology laboratories. The weighted overall accuracy in each blind test set was >63%, with a combined accuracy of 64.3% across both viable and non-viable cell lines, demonstrating model robustness and generalizability beyond the result expected from chance. Distributions of predictions showed clear separation of correctly and incorrectly classified cell lines. Binary comparison of viable/non-viable embryo classification demonstrated an improvement of 21.9% over cell lines accuracy (P = 0.042, n = 2, Student's t-test), and SOP procedure of QC comparison demonstrated an improvement of 42.0% over embryologists (P = 0.026, n = 2, Student's t-test). The superior accuracy of the Life Cell AI UAB model could lead to improved quality control assessments of samples in Biobank. It could also assist in standardizing QC methods of cell lines and stem cells across multiple environments while eliminating the need for complex time-lapse imaging equipment.