Routine Assessment of Fluorescence Microscope Performance
An ImageJ Macro to Speed Up PSF Display and Analysis
- Figure 1A: The macro generates a stack of 2 images. The first plane displays the projection in Z of the acquired stack of images, as well as the XZ and YZ planes crossing the center of the bead. It also shows the date of acquisition, the FWHM in XY and Z extracted from the Gaussian fit of the bead signal, as well as theoretical values for FWHM generated with the “PSF generator” ImageJ plugin.
- Figure 1B: The macro generates a stack of 2 images. The second plane shows pixel intensities along lines crossing the bead in its center respectively in the XY (top) and Z (below) dimensions. The points are fit with a Gaussian function (continuous line in the graph) and FWHM values are extracted. The stack name contains information about date of acquisition, microscope, objective magnification and numerical aperture, presence of an optovar, and measured FWHM values.
- Figure 2A: A scratch on the front lens of the objective generating an asymmetric bead image profile.
- Figure 2B: A scratch on the front lens of the objective generating an asymmetric bead image profile. (B) New objective with same specifications as in Figure 2A.
- Figure 3A: Distortions due to optical components other than the objective. Image acquisition with a misaligned ‘Dual-Cam’ beamsplitter (MAG Biosystems).
- Figure 3B: Distortions due to optical components other than the objective. Image acquisition without the misaligned ‘Dual-Cam’ present in Figure 3A.
Modern automated fluorescence microscopes are complex devices, dedicated to produce accurate representations of objects at a resolution exceeding the resolving power of the human visual system up to about 1000-fold. The fidelity and efficiency of this imaging process critically depends on the accurate ensemble performance of various optic and electronic devices. As the quality of the final image does not only depend on the performance of the microscope, but also on properties of the specimen, small performance flaws are often difficult to spot directly in any microscope image.
The use of the microscope response to sub-resolution light sources is a well established method for the assessment of instrument performance. Displaying and analyzing images, extracting relevant properties, comparing measurements to theoretical values and monitoring changes in the images and parameters over time though are laborious tasks. The automation of these tasks allows to dramatically reduce the workload and to apply the method frequently.
With the aim to distribute a procedure for a standardized equipment quality assessment within a community of partner sites, we chose the public domain image-processing software ImageJ  and assembled a set of semi-automatic image processing routines. The macro generates orthogonal projections from bead images along the lateral and axial dimensions which are displayed using a customized look-up-table to color code intensities. A Gaussian curve is fit to the intensity profile of a fluorescent bead image and full-with-at-half-maximum (FWHM) values are extracted. FWHM values of ideal bead images for various optical conditions were generated using the ‘PSF generator' or the HuygensPro software  and were stored within the macro. These values are displayed alongside the measured FWHM values to give a reference for the relative performance quality of the equipment (Fig.1).
Over a period of several months, we have analyzed the performance of four heavily used upright and inverted microscope setups. Our systematic study revealed that, objectives mainly suffered from careless use of automated microscope stages, particularly when different stage-inserts and sample-holders are used.
Front lenses of objectives are indeed very sensitive to scratches and hits, and can very easily be irreversibly damaged. This implies that, at sites where microscopes are heavily used by a group of customers with largely different experimental questions and professional skills -like in microscopy facilities-, it is necessary to assess the performance for objective lenses regularly. We found that checking performance at a weekly interval was a good compromise between workload and reaction time to spot performance flaws on systems that are used for more than 1500 hours per year.
The routine assessment also enabled us to immediately spot a number of performance shortcomings not originating from the damage of objective lenses, like a defective beam splitter (Fig. 3), vibrations generated from a broken camera fan, uneven illumination of the back focal plane in a spinning disc confocal microscope and fidelity flaws of a piezo stage.
The macro has been distributed to several scientific institutions in Europe and in the US, and has been well received as a tool in a number of imaging centers. A site list is available upon request.
Other procedures to test the performance of microscopes have been presented earlier, for example through generating Sectioned Imaging Property (SIP)-charts . The SIP-chart procedure reveals information about the axial resolving power of a microscope for each position inside the field of view and about the uniformity of field illumination alongside several other very useful parameters. It does not allow assessment of the lateral resolving power though. The preparation of suitable samples for the SIP charts is quite demanding, so we relied on kind gifts from the Brakenhoff lab. The analysis of the images is web based, requiring to load large sets of image data over the internet. The SIP-chart method is suited to assess sectioning, but not conventional microscope performance. The routine presented here complements the SIP chart method; it makes use of easy-to-prepare samples and reveals information about the apparent lateral and axial resolving power. The macro runs on any desktop computer and the method applies to both sectioning and non-sectioning systems. It does not provide information about the axial resolving power of a microscope over its entire field of view though or about the uniformity of field illumination.
The macro is freely available from the authors upon request. We appreciate any feedback on its performance.
Sample Preparation and Acquisition of Fluorescent Bead Images:
1. Dilute the purchased fluorescent beads stock solution 1:10'000 in water.
2. Deposit small droplets (5-10 μl) of this suspension on a high-quality 24x24mm cover glass.
3. When the droplets have dried, deposit 10μl-droplets of embedding medium on top of the beads.
4. Mount the cover glass with the embedded beads on a standard 26x70mm glass object carrier.
5. After 24 hours at room temperature, seal the cover glass with colorless, transparent nail polish.
6. On scanning microscopes, set the number of scanned pixels per image to 256x256, and zoom in accordingly to generate a lateral pixel size of 60-70nm. To reveal the full z-shape of the bead image, fully open the detection pinhole. On wide-field systems, pixel sizes will vary depending on objective magnification and camera chip pixel size. The area captured should be at least 15X15x20micrometer(x,y,z) in size. Note that the goal here is not to match Nyquist criteria but to find a good compromise between accuracy and speed.
7. For all microscopes, irrespective of objective magnification or numerical aperture, record z-stacks containing 100 images spaced by 200nm of a single, fluorescent bead. Make sure the emission of neighboring sources is not interfering and that the bead is roughly centered for all (x,y and z) dimensions.
Processing the Acquired Data:
1. Copy the macro ‘MIPs for PSFs all microscopes V13_.txt' into the ‘plugins' folder of ImageJ. The macro is freely available from the authors upon request [information on how to work with macros is available from the imageJ website at http://rsb.info.nih.gov/ij/developer/macro/macros.html].
2. Open a stack of images representing one fluorescent bead in ImageJ. Optional: use the LOCI-import plugin for data formats not available from the ImageJ menu.
3. Run the macro MIPs for PSFs all microscopes V13_ from the ‘plugins' menu.
4. Enter metadata information about date, microscope, magnification, NA of the objective, and pixel size as prompted by the macro user interface and press ‘enter'.
The macro will then automatically,
• Find the brightest pixel in the stack, interpreted as the center of the fluorescent bead image.
• Crop the stack axially to 20 μm with the fluorescent bead in the middle. Fill in missing planes if necessary.
• Crop the stack laterally to an area of 15x15μm centered around the fluorescent bead.
• Generate a maximum intensity projection, and re-slice in XZ and YZ through the brightest pixel, generating orthogonal views of the bead image.
• Assemble three images and resize the picture, so that a field of view of 15x15μm is represented within the same area at the same size on the screen.
• Plot pixel intensities along straight lines crossing the center of the bead image in the XY and Z axis.
• Fit the intensity data with a Gaussian curve and extracts the FWHM.
• Generate a picture with a new name figuring all metadata information plus the values of the FWHM. It can be automatically saved as well. The dates as well as the calculated and theoretical FWHM values are written into the picture. Note that theoretical values are calculated for a wide-field microscope by generating first a theoretical PSF with the "PSF generator" ImageJ plugin .
• Generate an output similar to figure 1.
5. (optional) The first images of the stacks containing the orthogonal cuts through the bead images can be assembled into a larger stack to easily spot changes over time, using the ‘stack sorter' plugin of ImageJ.
For the acquisition of three-dimensional images we use green fluorescent beads (Speck microscope point source kit, Invitrogen P-7220) of a diameter of 170nm.
The High-quality cover glasss used here are specified 0.17 ± 0.01 mm, Hecht-Assistant #1014/2222.
Embedding medium is ProLong Gold Antifade Reagent (Invitrogen, P36930).
ImageJ is a public domain Java image processing program available from: http://rsb.info.nih.gov/ij/; distributions are available for Windows, Mac OS, Mac OS X and Linux operating systems.
The macro ‘MIPs for PSFs all microscopes V13_.txt' is available from the authors upon email request (use Email button below)
Information on how to work with macros is available from the imageJ website at http://rsb.info.nih.gov/ij/developer/macro/macros.html
Optional: The LOCI-import plugin is available at http://www.loci.wisc.edu/bio-formats/imagej
We thank Dominique Spirig for the patient recording of bead images over weeks, as well as Karin Aumayr, Stefan Terjung and Timo Zimmermann for critical reading of the manuscript.
2. Griffa, A. et al.,: Comparison of Deconvolution Software in Microscopy - A User Point of view, Part I and Part II, Imaging and Microscopy, vol.12, issue 1+2, 2010 (accepted) and http://bigwww.epfl.ch/deconvolution/; Huygens Professional, Scientific Volume Imaging SVI http://www.svi.nl/products/professional/
3. Brakenhoff, G.J., et al.,: Characterization of sectioning fluorescence microscopy with thin uniform fluorescent layers: Sectioned Imaging Property or SIPcharts. J Microsc, 2005. 219(Pt 3): p. 122-32.