Computational Video Bioinformatics for Understanding the Dynamics of Living Cells

Recent advances in microscopy technologies such as high-throughput imaging, super-resolution, and 3D microscopy have revolutionized our ability to study cells and their underlying biological processes. However, many research groups are overwhelmed by the quantity and complexity of this new data. Traditional methods are time consuming, subject to bias, and difficult to reproduce. Because of this, it is highly advantageous to develop convenient software and tools to help cell researchers perform and analyze experiments. Video bioinformatics is an interdisciplinary field that automatically processes, analyzes, and visualizes biological spatiotemporal data using biology, computer science, and engineering methods. Here we present three video bioinformatics projects and software toolkits that automatically analyze, classify, and visualize biological processes and structures in multidimensional image sets. All three software packages were developed using novel machine learning, image processing, and computer vision algorithms. Unique microscopy datasets were collected for each experiment and were used to test and validate each developed software package. (1) StemCellQC, a bioinformatics toolkit that can automatically extract features from phase contrast videos of human embryonic stem cells, produce analyses, and classify cell health. (2) PhaserR4D, a software that can produce live 3D phase contrast videos by fusing phase and fluorescent image stacks captured on commercially available microscopes. (3) DendritePA, a pattern recognition software that can analyze subpixel protein trafficking events in neurons by using spatiotemporal information present in multichannel fluorescence videos