Life scientists routinely use microscopic images to understand the rules that drive a set of individual cells to develop into an organism. We aid in such research efforts by custom-building microscopes able to image volumes of unprecedented scale, deeper inside tissues, and samples developing over long time. We work in close collaboration with a number of biology labs to record developing Drosophila, Tribolium, Volvox, and Danio rerio, as well as fixed mouse tissues, among other model organisms. We face a range of challenges that include samples moving out of the field of view, light-induced phototoxicity, spatio-temporal limitations, and transferring, storing and analyzing extremely large datasets. To overcome such issues, we use custom microscopes that exploit the tight integration of state-of-the-art hardware and intelligent algorithms to implement, for example, adaptive optics, adaptive control, and statistical inference like deep learning. Our expertise is focused on multi-view, multi-sample lightsheet microscopy; confocal spinning-disc microscopy; and two-photon mesoscopy.
Volumetric imaging of biological systems across multiple spatial scales is pivotal to understand how tissues, organs, and organisms emerge from a collection of molecules and cells. In order to look at biology across scales, imaging tools must not only be able to resolve the smallest structures of interest, but also span extensive lengths at the highest possible resolution. Imaging large samples poses several key challenges to data acquisition, processing, and quantitative analysis. For example, thick and opaque samples require methods capable of imaging deep inside the tissue. Large samples need fast imaging modalities as well as high-performance data handling and processing. Accurate segmentation of subcellular structures requires high-resolution images. To overcome these challenges, we have built a high-throughput, two-photon mesoscope capable of imaging large, fixed tissues as big as one cubic centimeter, in-situ, and with sub-micron resolution. We aim to use these novel technologies to leverage our understanding of tissue architecture and self-organization.
As a particular application of our meso-imaging platform, we aim to digitally reconstruct an entire mouse liver. We work in close collaboration with biologists in the Zerial Lab at the MPI-CBG and theorists at the MPI-PKS to perform large-scale analytics on the tissue such as statistics on the cell localization, size, polarity, density, and connectivity, to aid the formulation of multi-scale models for systems biology.
Imaging samples that develop over long times requires smart microscopes taking routine tasks, like adapting the microscope, off the shoulders of the operator. Smart microscope automation enables reproducible, standardized imaging of living samples and thus, systematic investigation of developmental processes at a new quality. In order to adjust imaging depending on what happens in front of the objective, our live imaging microscopes analyze images in real time, while the recording is still ongoing. We implement our algorithms on top of general-purpose graphics processing units, using frameworks like OpenCL, Keras and Tensorflow.
We build microscopes that can detect samples in space, follow them over time and adapt to keep the samples in the field of view and in focus. They can autonomously decide which samples in a collection to image in more detail, and which samples to discard. Furthermore, they adjust temporal resolution to observe samples at the right time with optimal frame rate, while saving light during developmental stages which are not of primary interest.
Color-coded maximum projection of H2B-YFP tagged Drosophila embryo undergoing gastrulation. Imaged with custom built lightsheet microscope by Debayan Saha.
24 hour timelapse of a developing Drosophila embryo, histone-RFP tagged. Imaged with the XWing microscope by Robert Haase.
Involved lab members
- Alexandr Dibrov
- Coleman Broaddus
- David Chen
- Debayan Saha
- Justina Stark
- Laurent Abouchar
- Kira Vinogradova
- Martin Weigert
- Mauricio Rocha Martins
- Robert Haase
- Uwe Schmidt
- Advanced Imaging Facility, MPI-CBG.
- Eaton lab, MPI-CBG.
- Jug lab, MPI-CBG and CSBD.
- Light Microscope Facility, MPI-CBG.
- Knust lab, MPI-CBG.
- Norden lab, MPI-CBG.
- Tomancak lab, MPI-CBG.
- Loïc Royer, CZ Biohub.
- Sbalzarini lab, MPI-CBG and CSBD.
- Scientific Computing Facility, MPI CBG.
- Zerial lab, MPI-CBG.