This protocol describes how to capture flat field correction images for performing shading correction of fluorescence images.
When you acquire a fluorescence image on a microscope, the fluorescence is not uniform over the field of view. Due to uneven illumination and detection, some regions of the image, typically the center, are brighter, and others are dimmer. In addition, there is an offset to the image - the camera does not read zero when no light hits it - and this can vary from position to position within the image. Mathematically, our measured image (Imeas) is related to the true image (Itrue) by Imeas = Dark + Itrue * Flat, where Dark is a dark image and Flat is a flat field image. We can invert this to get Itrue = (Imeas - Dark) / Flat. This correction procedure is implemented in the Micro-manager MultiChannelShading plugin, among other places.
To acquire flat-field images, we need a uniform fluorescent sample. We have found that the best results for flat-fielding of widefield images result from using concentrated dye solutions, as described by Michael Model (1, 2). In particular we use the following solutions on inexpensive dyes (links go to suppliers):
DAPI channel: 50 mg/ml 7-diethylamino 4-methyl coumarin in DMSO
FITC channel: 100 mg/ml fluorecein in water
Cy3 channel: 100 mg/ml rose bengal in water
Cy5 channel: 100 mg/ml acid blue 9 in water
When you make these solutions, it's best to centrifuge them and filter them through an 0.22 μm filter to remove particulate matter than may make the fluorescence non-uniform.
Data Acquisition and Analysis
For best results, you should acquire separate flat-field images for each objective and fluorescence channel. In our experience, these images are pretty stable if you don't realign your microscope, so you shouldn't have to do it that frequently.
These instructions assume you're using ImageJ for your analysis. If you're not, it should be straightforward to adapt them to another program.
First, you must acquire a dark image. This is acquired with no light reaching the camera. To acquire it:
Set up the microscope so that no light reaches the camera. Typically, I set the microscope to use the eyepiece light path.
Set the camera to a short exposure time (10 - 100 msec; the exact value is not crucial).
Acquire a time lapse of 100 - 500 images
Calculate the average of these images. This is most easily done by opening the image stack in ImageJ and using Image → Stacks → Z-project. Select average intensity for projection type.
Save the resulting average image. This is your dark image.
Second, you must acquire flat-field images. To do so:
Prepare a slide of an appropriate flat-field solution by placing a drop between a slide and a coverslip.
Mount the slide on the microscope and focus on the solution. If there is an air bubble in the solution, it's easy to focus on that; otherwise you can use the edge of the coverslip to focus on.
Move the microscope objective to a uniform area of the dye, away from bubbles and and the coverslip edge.
Acquire many images of the sample in different positions. I typically acquire a 9×9 grid of images overlapping by 50%. This allows us to average out non-uniformities in the dye solution.
Load the resulting image stack into ImageJ, and use the Z-projection tool to calculate a median intensity projection. Choosing the median intensity makes this robust to outliers, such as fluorescent dust particles. Important: the resulting projection will be a 32-bit image.
To convert it back to 16-bit, make sure that Edit → Options → Conversions → Scale when converting is unchecked. If not, ImageJ will change your intensity values when you do the conversion. Use Image → Type → 16-bit to convert back to 16-bit.
Subtract off the dark image. Load the dark image, if it's not already loaded, and use Process → Image Calculator to subtract the dark image from your median flatfield image.
The resulting image is your flatfield image. Save it as a TIFF (or other lossless file format).