Removing Bottlenecks Between Image Acquisition To Analysis
Lattice light-sheet microscopy, a rapid volumetric imaging technique, acquires images using a plane of excitation which results in images being acquired at an oblique angle. For a typical acquisition of two colour movies of endosomal trafficking at 4 volumes per second for 20 min, about 20 TB of data is generated. Visualisation of the data to adjudge the quality and proceed to analysis is the first bottleneck. This involves deskewing of the data and rapid visualisation. Visualization will be performed on a custom visualiser or using virtual reality (VR) The second bottleneck is image segmentation. While routines like water-shed segmentation can perform reasonably well, they are slow and computationally intensive. To speed this process up, machine learning based approaches will be employed. However, for this, training data sets that accurately capture the signal containing voxels. Post segmentation, curating the data is a cumbersome process. I plan to utilize VR with direct gesture controlled functions to swiftly generate training data set based on semi-manual approaches. Lastly, complete visualisation of all steps, including further analysis such as tracking and motility analysis will also be integrated in VR.