SpinX: Time-resolved 3D Analysis of Spindle Dynamics using Deep Learning Techniques

Session

Stream 2: Machine Learning for Image Analysis

Authors

David Dang (1, 2), Christoforos Efstathiou (1), Dr Dijue Sun (1, 3)

Affiliations

1. School of Biological and Chemical Sciences, Queen Mary University of London
2. Department of Informatics, King's College London
3. Wellcome Trust Sanger Institute

Keywords

Deep Learning, 3D modelling, image analysis

Abstract text

Live-cell movies generate terabytes of data. However, manual analysis of this data is prone to error and can easily exhaust days of research time, thus limiting the insights that can be gleaned from high-resolution live-cell microscopy. Manual and automated analysis has been hard because of discontinuities between the distinct frames of 3D live-cell movies. 

We present SpinX, a comprehensive and extensible computational framework which bridges the gaps between discontinuous frames in time lapse movies by utilising state-of-the-art Deep Learning technologies and modelling for 3D reconstruction of highly mobile subcellular structures. Using SpinX, we are now in a position to precisely track and analyse the movements of multiple subcellular structures within minutes, including the cell cortex and mitotic spindle. We demonstrate the utility of SpinX by employing it to define the precise roles of spindle movement regulators that ultimately determine the plane of cell division. We illustrate the extensibility of SpinX by showing how it can also be used to infer the regulation of complex cortex-microtubule interactions. Thus, SpinX provides an exciting opportunity to study spindle dynamics in relation to the cell cortex using hundreds of time-resolved 3D movies in a sophisticated way.

References