Panoramic projections of life
Max Planck researchers process microscopy images in real time
Cutting-edge microscopes produce fascinating images, with an ever-increasing speed and resolution. The side-effect of this advance: Storing and processing data becomes more and more a challenge since vast amounts of data accumulate - a microscope of the latest generation can easily generate 25 terabytes a day. Researchers at the Max Planck Institute of Molecular Cell Biology and Genetics (MPI-CBG) in Dresden now trained their microscopes to process the images whilst taking them. So it's the results, not the raw data, that are stored. This reduces the amount of data enormously and makes the further analysis way easier.
"Unfortunately, microscopes are rather stupid for a start", says Jan Huisken, group leader at the MPI-CBG, who managed the project. Almost stubbornly the devices take large rectangular images, regardless of whether or not the images contain relevant information, and having no idea where the sample is. This results in a huge three-dimensional data block, an image of the sample – huge data that deploy massive storage and computing power, but containing a lot of information that nobody needs. Benjamin Schmid, postdoc in the Huisken lab, wanted to change exactly this. So he trained the microscope prior to the experiment to understand what is important and what is not. And now here they are with an intelligent microscope: It scans for relevant data, records the development of multiple embryos in parallel and projects a map of all labeled cells for each embryo in less than ten seconds.
Processing data on-the-fly and storing the result, not the raw data, makes a big difference concerning the data volume: The Dresden research team achieved a 240-fold reduction in data rate. This is a big step forward, actually the step back from terabytes to gigabytes. „This will enable biologists to perform a number of experiments in parallel – this is what they need and what is indispensable for image based systems biology”, says Huisken. His team managed to observe the development of 12 zebrafish embryos in parallel with light-sheet microscopy (SPIM). From the data, a map of all labeled cells is obtained for each embryo in real time. Taking insights from cartography, 2D map projections are generated - panoramic projections of life. The information form all 12 experiments is finally merged in a single image revealing characteristic migration patterns of the cells on the embryos’ surface.
The recent work is a proof of principle: The existing limitations are now overcome, processing and compressing raw image data in real time works and is extremely efficient. The next step will be to apply this method to other model organisms with different characteristics.