- Measure the point spread function of the 20.309 fluorescence microscope
- Examine the effect of NA, magnification, and optical corrections on PSF
- Generate a dataset for deconvolution problem set
Materials and methods
Exercise left to the reader
Python Libraries / Source
Microscope at Station 9
Built and tested. Images bright field and fluorescence at all magnifications. Code available for controlling pico motor stage for PSF images.
Microscope at Station 5
Feb 11 2011
The microscope currently at station 5 is not finalized. Bright field images reveal the sample along with a ghost image of the sample due to some reflection. One possible reason is that the 200mm imaging lens is too far from the objective.
Feb 12 2011
The ghost reflections disappeared when decreasing the bright field illumination. The microscope exhibits some astigmatism apparent when entering and leaving a focal plane. It is most likely due to a dirty lens in the imaging path.
Microscope at Station 3
Feb 14 2011
This microscope (formerly at station 8) has a complete optical construction, and was utilized to obtain images down to the 100X objective without major issues.
The computer controlled stage is being attempted with Matlab software, but it is really easy to just switch to Python control if preferred. The only caveat would be that the computer at station 3 doesn't have Python yet (ask (Steve)^2)
The camera adapter for Matlab has been successfully installed, but again it would be straightforward to switch to Python control based on the code developed by other class members.
ImageJ Tutorial: Deconvolution Lab
Installing and Running Deconvolution Lab Plugin:
1. Download and install ImageJ
2. Go to the deconvolution lab website Deconvolution Lab
- Scroll down to 'Download and install' then right-click and save 'DeconvolutioLab.zip' into C:\Program Files\ImageJ\plugins or for Macs Applications/ImageJ/plugins. Leave it unzipped, ImageJ will automatically detect and install it.
- Open ImageJ, go to the Plugins menu and 'Deconvolution Lab' should be listed
3. Download tutorial image stacks (Spring 2014 students, collect a "3D PSF" data set from your fellow classmates and follow the 3.1. - 3.2. set of instructions)
- Go back to Deconvolution Lab.
- Scroll down to 'Examples' (right before the sample images) and go to Deconvolution dataset. Here you will find 3 example image stacks and their PSFs which you can use to test the 'Deconvolution Lab' plugin (on the left column you can find the three Test Datasets as well as PSF Generator plugin.)
- Download the image stacks into whatever folder you want to open them from (your choice). The images will be a folder of TIF files taken at equal z-steps.
- 3.1. Convert your stack of "3D PSF" images into an image sequence acceptable by ImageJ (list of .TIFF images).
- You can use for instance the WriteNumberedImageSequence.m Matlab code below.
function WriteNumberedImageSequence (ImageStack, RootName)
for ii = 1 : size ( ImageStack, 3 )
imwrite( uint16( 65536 * ImageStack(:,:,ii) ), [RootName, ' ',num2str( ii ), '.tiff'] );
- 3.2. Have the image sequence grouped into a folder F1.
4. Import image stacks in ImageJ
- You will need to import the 'real' image stack and the PSF stack for the deconvolution lab.
- Go to File > Import > Image Sequence
- Highlight and open your image sequence from F1. Click 'ok' and ImageJ should show a slideshow of the image stack. You can view the stack as a 3D figure by going to Plugins > 3D > 3D Viewer.
- From any frame of the original image stack, crop a region of interest including a single PSF bead.
- Save the cropped PSF stack of images via File\ Save As…\ Image Sequence in F2.
5. Run Deconvolution Lab
- After opening the 'real' image stack (from F1) and the PSF stack (from F2), click on the 'real' image stack window and then go to Plugins > DeconvolutionLab. The Deconvolution Lab GUI will pop up. There are many parameters to give you the best deconvolution but here are the basics parameters to set
- Set the Algorithm to 'Richardson Lucy', which is a deconvolution algorithm (http://en.wikipedia.org/wiki/Richardson%E2%80%93Lucy_deconvolution).
- PSF image should be PSF image stack. Also, check the box next to 'Flip PSF quadrants(activate if PSF is centered)'.
- If you have a 'theoretical' image, you can compute the Signal-to-Error Ratio (SER).
- THEN, hit run and let the deconvolution begin.
The result will be the 'deconvoluted' image (as close to 'theoretical' image) of the 'real' image stack using the PSF stack. The 'deconvoluted' image will still have some high frequency noise or what looks like fluorescent bleeding (white blur around object).
Creating a Theoretical PSF
- Open ImageJ
- Under Plugins, select "Diffraction PSF 3D"
- Enter parameters:
- Index of Refraction of the media = 1.398
- Numerical Apeture n*sin(theta) = .65
- Wavelength (perhaps in nm) = Nile Red (550-750) – 590 nm lower barrier filter cut-off, 640 nm center wavelength
- Longditudinal Spherical Aberration = 0
- Image pixel spacing = 185 nm
- Slice spacing = 156 nm
- Width = 32 pixels
- Height = 32 pixels
- Depth = 80 slices
richardson-lucy for 5 iterations - http://web.mit.edu/sivakami/Public/20.345-lab1/richardson-lucy%20for%205%20iterations/
richardson-lucy with generated psf - http://web.mit.edu/sivakami/Public/20.345-lab1/richardson-lucy%20with%20generated%20psf/
A Z stack of images were taken of rat intestinal cells using fluorescence microscopy. An image of an 190nm fluorescent bead wastaken. The bead image is used to approximate the three-dimensional point spread function. The deconvolve the original image with the PSF approximation, the Richardson-Lucy algorithm was used. The resulting image stack showed increased resolution, especially at the high and low values of Z (far above and below the focal plane). These images are stored on my public page
Intestinal cells: cells.tif
Deconvolved Image: Richardson-Lucy_of_cells.tif
Bead image: 3D_PSF_final2.tif
Fluorescence Microscopy in 3D (Paper)
EPA Deconvolution Lab
Pololu Stepper Motor Driver
ThorLabs DRV001 Stepper Motor Drive
ThorLabs NanoMax Stages