Tricks for Volume View and Deconvolution

Here are a few simple guidelines for deconvolution and presentation of deconvolved data. I’m posting them because it seems like I’m always forgetting one step or another, which results in waiting another 20 minutes for decon data to re-process…

Acquisition:

1. Remember that the larger signal to noise you can get will deconvolve better. Large numeric spread between the signal and noise gives the deconvolution software more room to work. This also observes that a 16 bit camera will acquire far better deconvolvable (is that a word?) data than a 12 bit camera. More numeric room = more deconvolution ability. So use a good camera AND take the longest exposure you can get away with! In some cases increasing binning to get more signal will produce a better final result even though you lose the XY resolution.

2. After acquiring your data make sure you save whatever you acquired. Deconvolution has a way of finding crash points in software…

3. The immediate next step is to subtract background. Any background in the image that is not present from convolution will show up as a weird halo around areas of signal. An example of removing background using Elements is below:

4. Next we crop the image. This will speed up processing time and reduce the potential for memory problems on 32 bit systems. If you’ve never deconvolved the data at all crop down to only a small portion of the object of interest. That way you can decon, adjust settings and re-decon without waiting for too long. ***Remember to crop in X,Y AND Z! Extra Z planes below and above the specimen will only increase processing time.

5. Save the cropped & background subtracted image!

6. Run the decon.

Viewing:

1. Most viewing systems have a volumetric and orthagonal slice viewer. In a nutshell the slice viewer is faster, shows the deconvolution’s effect quickly but isn’t as visually impactful. It essentially 2-d single-axis views, (X/Y, Y/Z, X/Z) drawn from 3-d data.

The volume viwer is the one most people like to see. So how do we get the best volume? Volumes represent voxels in XY and Z space. The voxel is colored with an intensity pulled from the source data, and when the user rotates the object the brightest voxel that shows up at any given angle of rotation is brought to the front of the image view. This means that a bright nucleus inside of a cell will show up easily even though the cytoplasm may also be bring and in front of the nucleus. This is the “Maximum Projection” view that is the standard.

The next view is either an isosurface or an alpha blend. An Isosurface first finds the object boundaries (think of the outside of a tennis ball) and draws a solid sheet over the boundary, sort of like the skin of an airship. You then view this shell. An alpha blend uses the object’s light absorption, which usually shows surfeaces better but includes internal signal. In the isosurface image a nucleus would be hidden inside of the cell cytoplasm, whereas in the alpha blend you would see a semi-transparent cytoplasm and a faint nuclear signal.

Each of the projections used rely on hardware graphics cards. Most cards have several setting options that can be controlled to adjust the image quality and processing inside the card. The default settings for most software packages on these cards is the middle performance range. In order to get the best view we need to crank these settings up to the maximum level possible. In Elements the settings can be found in the image window. In other applications you can find them in preferences windows or in the 3-D viewer window. Here’s an example of these settings

After setting up your 3-D settings, make sure to check any options allowing for “recalcuate Slices in Z” – these can help fill in the “stack of pancake’ look found on undersampled data. Finally reset your image scaling (LUT’s) to show the signal and not the background. With some adjustment you should have a fine example of your 3-D data!

– Austin