This afternoon I took a break from a long editing job I've been doing and took a few pictures and videos of living bacteria in a Vaseline prep.

I used the white LED and a very blue (maybe 'Royal' Blue) Leitz filter. The light was lined up with a lower power objective (10X) and so the phase ring (40) is not perfectly centred -- but this doesn't matter, see picture 1908. The image 1909 was made with a Lomo Phase 40/0.65; Lomo Phase condenser 0.9; mag changer at 1.6X; 12.5X Periplan eyepiece and the Quickcam Pro 4000 with 12mm Marshall lens. Total magnification on the microscope monitor (Sony E450) was 2790X.

 Link : ../PICTURE/1908.jpg
 Link : ../PICTURE/1909.jpg

To get even illumination the field diaphragm is closed as far as it will go and the pattern is centred with the mirror tilt screws -- see 1908. To do this a lower magnification objective is necessary. The LEDs are perfectly centred on the mount, but the whole assembly can be moved laterally + - 1.5mm in any direction as well. This adjustment was needed for the original filament lamps.

Don W

Interesting images Don. I presume the main point you are making is about necessary lighting adjustments.

Are the dark spots inclusions in the bacteria?
Also, was there a particular reason for selecting a blue filter as against green?

Selwyn

I have a selection of pictures with different filters. This one showed more internal structure than the others.

I will have more time soon -- I hope -- to take more and post a selection. I'll also try a Royal Blue LED without filters. I had hoped to get UV LEDs fitted with pins and diodes, but am far too busy right now. Also the voltmeter of the regulated power supply was damaged in transit and I need to find a replacement -- but I've been exceptionally busy.

Don W

Hi again,
I have just looked at all the pictures I made yesterday and another dozen or so made a few minutes ago. I used the phase contrast green filter, a daylight correction as supplied by Zeiss and Leitz and another less blue Leitz filter. None, not a single one, comes anywhere near the picture I posted yesterday (1909) for detail of the internal structure and most of them are not as sharp. None are worth posting. Everything was the same and I used the same field as I had not moved the stage. Again I fear I am pressed for time so this is all I can do for the moment.

Don W

Hi Don,
As always, your shared images and comments are very much appreciated.

A general question: how do you feel the white LED illuminator compares with your optimized 100W halogen setup?

Regards, Keith

The white LED seems to give me as much light as I need and there is no flicker in the video clips.

I'm wondering now if my plans to use other colours will be a waste of time? Certainly the Royal Blue LED performs well and I know there are some advantages to using UV. But I might be wasting time and money if I bother with green and other shades of blue.

Another nice thing seems to be the absence of hot spots. I have removed the two IR blocking filters I had in the light path as the LEDs don't seem to emit much in the IR region. But I haven't checked this aspect thoroughly yet.

Don W

"scitech200" wrote:
> A general question: how do you feel the white LED illuminator compares
> with your optimized 100W halogen setup?

I found a little time to test a white LED (LXHL-PR02) with video.
I've posted a clip of bacteria at the Vaseline seal of a prep about three days old:
 .../EDFWilliams/VIDEO/912-3DX.avi

There is plenty of light for doing this job and I used the highest possible magnification the microscope will give with a (40X objective) for the last half: 12.5X Periplan photo eyepiece; 40X Lomo Phase objective; condenser Lomo phase; Mag changer set to 1.6X and 2.5X; camera QC Pro 4000 with 12mm Marshall lens.

As you can see I used bright field being and was just too lazy to change to BF objectives. Quite a bit of detail is visible in the cells and thick capsules are also visible.

Don W

I took stills at the same time as the videos; so the data is the same except the stills are 1280 x 960 in size. I did no sharpening but did stretch the histogram to fit between 0 and 255 in Photoshop -- for each picture:
 Link - Photo index
They start at 1980.jpg and go on for a while.

Quite a lot of detail can be seen. Most of us will benefit by the large magnification. But children and maybe teenagers would see as much with no extra blowing up. I repeat -- I used no filters, just the white LED. Some of the images show a lot of internal detail and capsules are visible.

Don W

I had some time and so I did a few tests with filters and the white LED. Everything is the same as for the previous pictures, but I used different filters.

The first are the untouched images as they come from the camera:
  ../LED/P2009
  ../LED/P2010
  ../LED/P2013
  ../LED/P2014
  ../LED/P2017

The next are the PS CS Histograms of these pictures:
  ../LED/H2009
  ../LED/H2010
  ../LED/H2013
  ../LED/H2014
  ../LED/H2017

And finally the same after stretching:
  ../LED/2009
  ../LED/2010a
  ../LED/2010b
  ../LED/2013
  ../LED/2014
  ../LED/2017
2010 was fiddled with a bit so there are two different degrees of stretch a and b.

Don W

Don,
I'm not familiar with the Photoshop operations you are using for the "histogram stretching" but is it accurate to consider it as a contrast enhancing, data processing technique?
To try and understand this technique, I took the example labeled as 2017 (blue-green filter?) and consolidated the available information on a webpage:
  ../MicGroup/Msg_08/DW_bacteria_01.htm
It would be great to also have the modified histogram available.
There is a significant enhancement of detail, but I'm wondering what part of the data processing causes the overall color changes.

Maybe I should go try Photoshop Elements 5.0 and see if I can reproduce something similar from your P2017.jpg data....

As the bacteria seem to be stationary for these tests, i.e. locations are apparently the same during filter changes, how did you prepare this sample? Or is it indeed the same vaseline slide you used for the most recent videos?

I know very little about bacteria, but these seem to be some form of the general rod-like type - bacilli? Also, as I recall, they are single celled organisms that have the cytoplasm (primarily water) contained within a membrane capsule that has a tail (flagellum) that is used for locomotion. Also, the DNA floats freely in the cytoplasm so there is no well defined nucleus as for protozoa and human cells. There are also clusters of ribosomes inside the cell that produce the basic proteins required by a developing cell, etc.
[ The bacteria dimensions are ~ 1um ]

So, what parts can we be expected to see in these enhanced images? For instance, what are the well defined "globules" inside the cell?

Thanks once again for sharing all of this interesting data with the group.

Regards, Keith

"Don Williams" wrote:
> I had some time and so I did a few tests with filters and the white LED.
> Everything is the same as for the previous pictures, but I used different filters.
> The first are the untouched images as they come from the camera:
> ...
> http://www.science-info.net/pages/EDFWilliams/LED/P2017.jpg

Don and the group,
I did a prelim test with Photoshop Elements 5.0 with your P2017.jpg image. It appears that I get an identical histogram displayed with the function: Enhance > Adjust Lighting > Levels. Then with the 'Auto' option the output image is very close to your processed image (2017.jpg)and it has a histogram as shown on the updated Archive webpage:
  ../MicGroup/Msg_08/DW_bacteria_01.htm

Now I have to find out from available Photoshop documentation what this 'Auto Levels' option does to an input image, versus the manual manipulations that are available. Do you (or other group members) have any input on this particular Photoshop operation?

Regards, Keith

"scitech200" wrote:
> Maybe I should go try Photoshop Elements and see if I can
> reproduce something similar from your P2017.jpg data...

Don,
I have played with one of your images using deconvolution in ImageJ and the auto levels option in Photoshop CS3. These are to be found on the following link:
  ../picasaweb.google.com/selwyn.stleger/P2009
Hover the cursor over the images to see the full captions. Click to enlarge images.

I should be interested to hear views on whether any of these particularly enhance visibility of structure.

Selwyn

I have added two more images to the folder:   .../selwyn.stleger/P2009

These apply the free blind deconvolution utility Deblur to Don's original image. The first application is with Deblur set to normal blur and the second with it set to severe blur. In both cases the utility was forced to concentrate on a region of the image with plenty of bacteria.

Does anyone know of other easily available blind deconvolution programs? It would be interesting to compare among them and with non-blind deconvolution.

Selwyn

Selwyn,
Today I have to do some hobby-not work, but a few quick comments:
- These data processing techniques are an interesting exercise, but I also asked about the structural features that can be expected to be observed for these particular bacteria (at this magnification).
- I do not have the expertise to comment on the value of the enhanced images versus the originals, so hopefully other group members will assist.
- I did some experimenting with Photoshop's levels adjustment feature. It appears that the auto-leveling works by implementing a histogram stretch for each of the R,G and B layers and then merging for an output image. In a case such as Don's P2017 image this can lead to significant 'color casting'.
Again, other members may have more input on this general technique - which is evidently used extensively by astronomers and scientists involved with remote sensor data.
- I seemed to get the "best" enhanced image by manipulating the blue layer (channel) via a manual level adjustment, and so was very interested to see your deconvolution with level adjustment experiment.

This morning, I did get the Archive website section for 'High Resolution - Illuminator Tests' updated (now 3 pages) with Don's recent posts re using LEDs: .../MicGroup/Msg_08/Msg_08.3.htm (this webpage).

Thanks once again for your input.

Regards, Keith

"A. S. St Leger" wrote:
> I should be interested to hear views on whether any of these
> particularly enhance visibility of structure.

Hi all,
To properly investigate the structure of these bacteria, such as it is, one would use an oil immersion objective to get the best possible image with which to start. These, taken with a dry phase objective are not ideal.

The experiment was more to see how the white LED behaved with filters. Manipulating the image with Photoshop to make more detail visible -- was 'by the way.'

There is barely enough light to use the binoculars comfortably with phase and the green filter. But the camera makes all the difference in the world. It intensifies the image as well as some of the expensive cameras we had in my lab in days of yore. I don't know how I'd be able to get on without it now.

To someone like me who paid $175 000 for a Hewlett Packard system 2000; $78 000 for a Optronics Photomation P1700 rotary scanner and $8000 for a Versatec electrostatic printer (and my staff spent countless hours interfacing the devices to each other), what can be achieved using PS CS and ImageJ is miraculous. The computer had 192 kb of memory and took all day to do a fft of a matrix of 128 x 128 pixels; mask generation and back transform. The hard disk was 15 megabytes in size and the removable platters were the size of very large dinner plates. This PC on the floor nearby, can do in the time it takes a finger to come back off the enter key, the same manipulations on 64 times as much data.

Don W.

Hi Selywn,
Would you please let us know where you found the Deblur software.

I could only find some freeware from John Costella called Unblur, along with general references to blind deconvolution.

On playing with Photoshop's histogram stretching you get to understand that microscope images are relatively difficult to enhance. Most times, the input data is sparse and hence the output histogram has gaps. But I had no idea as to how versatile this technique is for "tonal adjustment" of a photograph, which I think we consider as a form of contrast enhancement with a photomicrograph.

Thanks, Keith

"A. S. St Leger" wrote:
> ...These apply the free blind deconvolution utility Deblur to Don's original image.

Hi Keith,
I inadvertently mislead you about the software. There is, I think, a commercial product called Deblur which I don't have.
I used something called Unshake which employs blind deconvolution. It is distributed free of charge and is excellent:
  www.hamangia.freeserve.co.uk

I should be interested to hear if anyone comes across other free or low cost blind deconvolution software. A general web search shows a lot of interest in this technique and I suspect it won't be long before open source stuff arises from microscopists (c.f. ImageJ) or astronomers.

However, it should be noted that the deconvolution software that presently plugs into ImageJ is non-blind and requires information about the optical system's characteristics and a few assumptions. Yet, it produces better focussed images which seem to show detail otherwise obscured. It's a toss up whether one prefers the ImageJ version or Unshake, but note that the latter does work on colour images.

With respect to your comment Keith I think Don's images are particularly sparse and thus a good test.

Selwyn.

Hi Selwyn,
Thank you very much for this reference!
It seems to have a lot of potential, per a "first look".
However, it not being a conventional Windows app may bother some members and so I put together some notes per my download and installation:
 ...MicGroup/Msg_08/Unshake_01.htm

I have just run a few tests with a sample of DonW's A.pellucida images and the results are quite impressive, in that Cahill's algorithms do not introduce many artifacts. I would encourage you to also test with some of your Ap images.

I think I understand the basic 2D FFT/deconvolution data processing, but I'm not sure about the app's 'amplitude' control. Do you have any input on this control?

If you are interested in seeing Cahill's UK patent, which includes a detailed description of his algorithms, let me know.

Best regards, Keith

"A. S. St Leger" wrote:
> ...I used something called Unshake which employs blind
> deconvolution. It is distributed free of charge and is excellent.