During the last decade microscopy has gone through a series of major improvements. The demand of microscopy technique has increased a lot and has brought the limits in optical resolution of light microscopes, as described by Ernst Abbe, to be extended by new technologies like 4Pi, STED [3], deconvolution and others. Scientists want to resolve small compartments and structures within a cell and but at the same time need to visualize a large field of view to be able to understand the complexity of biological organisms. Given the fact that optical resolution has been steadily improving up to a factor of 2-3x and that acquisition systems can acquire in a short timeframe huge 3D and even 4D stacks, the need to bring all this data together has become a major task of scientific labs. Recent developments in image formats, image management and data backup have been addressed by the imaging community but still needs to improve a lot to be useful in scientific labs.

FMI Core Facility Idea

Three years ago, the Friedrich Miescher Institut (FMI), which is part of Novartis Research Foundation, started a center facility for microscopy whose goal was to support scientists in both microscopy and imaging. FMI has three main areas of interest: growth control, epigenetics and neurobiology. In all fields the demand of microscopy has been increasing over the last years and will intensify even further in the future.
In some fields the scale of the structures of interest is at the limit of microscopical resolution (fig. 1), while in other areas one works at much larger scales, e.g. at the level of a whole microscope slice of the mouse brain (fig. 2). In order to cover the whole spectrum of possible applications, we have started to setup equipments for both Micro and Macro scales. Due to several reasons we have been focusing on light microscopy and have not set up other techniques like Atomic Force Microscopy, Electron Microscopy or Tomography.

Overcoming Abbe Resolution

In the light microscopy there are nowadays several approaches to overcome the Abbe Resolution, like structured illumination, STED (STimulated Emission Depletion, which works by quenching the side lobes of the point-spread function), and 4Pi microscopy.

With these new techniques resolution can be pushed down to an axial resolution of around 100 nm. Some other techniques, like PALM, can achieve even higher resolutions but are not applicable to time-lapse imaging. In parallel to these efforts, improvements in image processing and restoration have pushed the limits of the optical systems even further. Restoration algorithms like deconvolution can achieve a lateral resolution of 150 nm in biological samples imaged with standard microscopy techniques.

Large Field Of View

For large scale visualisation, there are in principle two possible approaches. Either high resolution pictures in 2D or 3D are acquired independently and later stitched together (mosaics), or techniques with large field of view like binoculars or macroscopy are used. The rationale of stitching is to allow the entire overview of the feature of interest at the highest possible resolution. Slide scanners are already widely used in pathology and histology, but start to be used as well in other areas like cell biology due to new fluorescence detection techniques. Recent developments in scanning tables allow driving motorized stages with average to good precision. If the geometric positioning is not enough accurate, stitching can be refined by calculating some form of correlation between the different field of views and refining the positions. This obviously requires some overlap of the different views. Robust software for large stitching is currently lacking. For the generation and handling of the large amounts of data generated in applications like stitching, not only hardware and software need to be adapted, but also new and better-suited file formats are required. Common file formats like TIFF, JPG or BMP fail to fulfill the needs of the scientist. Mosaic images can easily achieve a size of 0.8-2 TeraBytes, and mechanisms to quickly address single planes of the mosaic are essential. One file format explicitly designed for the handling of large data and the associated metadata is HDF5. Moreover, visualization tools must be adapted to allow the smooth navigation in these reconstructed, large specimens. Even though techniques in light microscopy are constantly improving, other techniques like electron microscopy or atomic force microscopy will always allow much higher resolutions. To address possible requests in this direction FMI decided to join a project with the Swiss Systems biology Initiative in Block Face Electron microscopy and tomography. This technique will allow viewing of biological samples in 3D at highest resolution. One can easily imagine how large a data stack will be for large field of views at this resolution. The final goal of this project is to combine light microscopy with electron microscopy and tomography and to start a Correlative Light Electron Microscopy (CLEM ) study. To localize specific structures one of the most interesting new developments in the field of CLEM techniques has been the use of quantum dots (QDot). This nanometer sized crystals of CdSe are brightly fluorescent and are more and more used in single molecule experiments and have high absorption in order to be detected in Electron Microscopy [1].