University of Sheffield (UK) news release, Feb.2007

If the image can be reconstructed from the diffraction in the subject itself it shouldn't be limited to X-rays. The news release is rather limited in what it says but we have some very clever minds among us. It may be enough for someone to figure out.

Gordon

Hi,
Maybe I'm missing something, but this sounds like X-ray diffraction in its current state of the practice. It is difficult to tell from the abstract.

An X-ray diffraction image is rendered in what is called "reciprocal space" and then must be resolved into real-space parameters, and then it is only a mathematical construct, which then must be further rendered into an image using other software. However I am also not aware of being able to do this on living cells, so perhaps that is the new innovation, and maybe their system creates a 3-D image all in one package.

It remains to be seen whether this is a quantum leap or whether, like DLC, it is a fresh coat of paint on an old house.

Don

Don,
I thought that too at first. But constructing a 3D image with no optical element to me means no pin holes etc.

This paragraph is sure not an old house with a new coat of paint. "/It is even possible to contemplate a solid-state optical microscope, built into a single chip with no optical elements at all. All the weakness, difficulties and costs of lenses would be therefore replaced by a combination of good quality detectors and computers/."

Nowhere does anyone claim to construct an image from DLC other than with an eye or camera. There is no connection between the two I can see other than the word diffraction.

I think this is a better description of what they are doing.
  Link : http://www.lbl.gov/Science-Articles/...

Gordon

"Gordon Couger" wrote:
> I thought that too at first. But constructing a 3D
> image with no optical element to me means no pin holes etc.

Howdy,
Current techniques in XRD, TEM, and SEM don't use optical elements per se, though they use magnetic lenses and apertures. The SEM and TEM are already solid-state microscopes, and most new diffractometers use solid-state detectors.

The second link you sent answered one of my questions:
"With transmission X-ray microscopy, samples are rapidly frozen and need no further chemical alteration or staining to be imaged." So they are indeed frozen.

My reference to DLC was an analogy--both are adaptations of current knowledge in their own way, though the "press releases" involved add a bit more sensationalism than might be warranted. I suppose I have an issue with the way some of these articles are written.

Don

Hi all,
They may be referring to the phase component of the generated image after Fourier transform, perhaps it is more of a computer image processing technique in the nature of phase contrast X-ray imaging, or digitally generated phase contrast light microscopy which is already being used, see the link below for the light microscopy info:
  Link : http://www.iatia.com.au/products/lifeSciences/...

regards, Matt

Don,
I may be interpreting things wrong. But when they talk of a single chip microscope with no optical components based on diffraction patterns using computers I think it worth considering the possibility. We don't use the power of computers very well. Computers get faster, faster then we can figure out how to make use of the speed. The astro photography folk make us look pretty bad.

Gordon

Hi Gordon,
You wrote:
> ...But when they talk of a single chip microscope with no optical
> components based on diffraction patterns using computers...

I had time this morning to take a look at your interesting links re Soft (low energy) X-ray Microscopy. I'm not sure about this 'single chip microscope' comment.

However, it seems that there are two distinctly different approaches being used:

1) Berkely Lab - ALS (advanced light source)  Mar.2004
The lead scientist has a unique background in cell biology and microscopy techniques and is promoting the idea of using a transmission X-ray microscope, similar in principle to a TEM. The published images are colorized renditions of 2D and 3D tomography data corresponding to X-ray absorption by the sample.
The optics of a light microscope are replaced with very sophisticated 'zone plates' that focus the X-ray beam at a pinhole adjacent to the specimen, which is flash frozen in milliseconds by a stream of cold helium. As DonH pointed out, this specimen preparation seems to be a key innovation that allows imaging of cells at spatial resolutions on the order of 25 nm (as compared with confocal LM at 200nm) and with magnifications of about 2500x.
The beauty of the overall technique is that the biological specimen is virtually undamaged and the complete image data is collected and processed in a few minutes.
It's interesting to note that the ALS beam has a flux density at least a million times higher than that for a conventional X-ray tube! We have to assume that this lab has access to state-of-the-art X-ray detectors and so I don't understand the significance of this high intensity beam. Regardless, as a National Lab available to cell biologists at other research labs, it's going to produce some outstanding images without the tiresome prep required for TEM/SEM work with the inevitable specimen damage.

2) Sheffield University (UK)  Feb.2007
Along with the SUNY group in the USA, they are developing a "lensless microscope". For any conventional microscope, they point out that "the tiniest error in the lens can make the waves interfere incorrectly, ruining the image. For this reason, a typical X-ray microscope image is about one hundred times more blurred than it should be."
The primary data is collected in a mannner similar to the classical Bragg X-ray scattering spectrometer used for studying crystalline structures. However, whereas an ordered crystal structure produces diffraction patterns that are relatively easy to interpret, cf. the contribution to the discovery of the DNA double helix, a biological specimen is way more complicated. Thus the innovation in this case relates to very sophisticated mathematical algorithms to untangle both phase and magnitude information required to produce a complete and accurate image.

I'm sure that other members will let us know if we are on track (or otherwise) with understanding this particular form of microscopy.

Regards, Keith

"scitech200" wrote:
> 2) The primary data is collected in a mannner similar to the classical
> Bragg X-ray scattering spectrometer used for studying crystalline structures.
> However, whereas an ordered crystal structure produces diffraction patterns
> that are relatively easy to interpret, cf. the contribution to the discovery
> of the DNA double helix, a biological specimen is way more complicated.

I think that about sums it up. X-ray diffraction can already elucidate crystal structures on the order of angstroms. I'm concluding it must be the software that's new--and quite an accomplishment it is!

Not to take anything away from their wonderful achievement, but the PR could use a little fine-tuning. But then again, PR is PR, and when someone accomplishes something like this, especially when your funding source wants to see results, then I suppose it seems necessary to make it dramatic.

DH

Someone earlier on this thread rightly mentioned something to the effect that diffraction phenomena using X-rays are no different in principle from those using light. Does this mean that similar software could be used to enhance light microscopy resolution (without lens')?
Presumably that would be useful to researchers who need living specimens rather than those frozen in time.

Selwyn

For those of you following the X-ray microscopy thread, here is the Rodenburg paper describing a "Lensless Hard-X-Ray Microscope":
  Link : http://www.kscitech.com/MicGroup/Msg_16/PhysRev.htm

Regards, Keith

"scitech200" wrote:
> ...here is the Rodenburg paper...

Ah! There we go. Thanks so much for that. I must admit that drawing is worth more than a thousand of my words.

It is disingenuous to call it a microscope, for sure.

Diffracting Don

Selwyn,
It's a bad weather day (early Summer) in the Northeast USA, so I spent some time on the computer.

Here is a Rosenburg paper (2006) that describes some "lensless LM" experiments:
  Link : http://www.kscitech.com/MicGroup/Msg_16/UltraMic.htm
utilizing the same algorithms that were applied to high resolution X-ray microscopy.

It's oriented toward lower magnification images with a wide field of view, but I think it may be an interesting read for you (and DonH, Gordon and other group members).

Regards, Keith

"A. S. St Leger" wrote:
> ...Does this mean that similar software could be used to enhance
> light microscopy resolution (without lens')?

Keith,
A bad day for fishing, a good day for microscopy.

This is taking the turn I hoped it would. The math is still beyond me. Maybe not, I was once told obvious meant something that could be figured out by someone that had enough understanding of the field to do it with a reasonable about of study and work, say, 8 or 10 weeks full time. Watch out if you get hooked up with a hydrologist. If I had a data set I think I might have 50/50 chance of doing what is described in the paper in the usual amount of time for me to get a major project done, a year or two with no interruptions if I can still concentrate like I use to be able to.

If it works better with shorter wavelengths I wonder how it would work with gamma ray using a pin hole source and moving the subject or the camera. I don't think there would be a diffraction pattern with gamma rays but, absorption or back scattered X-rays instead.

For us UV might be the best most of us can do and not run afoul of the federal government on working with radiation. Although getting a license for X-rays is not too hard.

Gordon

"Gordon Couger" wrote:
> If it works better with shorter wavelengths I wonder how it would work
> with gamma ray using a pin hole source and moving the subject or the camera.

I can answer that one with certainty. Gamma rays have a different character. We learned in X-ray analysis class that they have too much energy and tend to pass through matter without interacting with it.

Thanks Keith, it is a fascinating article.
Like Gordon I only follow the gist of it but it is a spur to study optical theory in greater depth.

It occurs to me from this and other recent developments that microscopy is going through a phase as exciting as that in the Victorian era. Sadly, this time much of this will be beyond the scope of amateurs but it is still interesting to keep abreast of developments. On the other hand some of the work on image processing from conventional optics is accessible (e.g. via ImagJ) and useful, it adds something to the fascination of microscopy as a hobby.

Conventional light microscopists can still delude themselves that they are seeing what is actually there, much as if they were viewing a macroscopic object unaided, even though what they are seeing is a contruct which, hopefully, has a one to one relationship with certain physical structures of their specimen. Realisation of the work of the Sheffield group and others will take the "seeing" out of microscopy. A construct ("image") resulting from the processing of data otherwise inaccessible to the human visual system will be displayed on a monitor with no direct visual equivalent on offer. Maybe that is not a bad thing because it forces people to recognise the artificiality of their constructions and hence consider their limitations.

If this work is taken forward so that it replaces conventional light microscopy and extends its range then maybe Moore's law will mean that my prognostication that it is beyond amateurs is wrong. It seems that the main cost is processing power. Perhaps every lab and many amateurs will have such equipment. There will still be romantics who wish to view their specimens directly.

Selwyn

"scitech200" wrote:
...Rosenburg paper (2006) that describes some "lensless LM" experiments

Hi Selwyn,
We hardly view the subject directly with a compound microscope. Most of the time we have to tweak the light a good deal to get what we want. We fiddle with the polarization and phase of the light to get more contrast or show things that otherwise are invisible. Almost every other field has gone a lot further with a computer than light microscopy. I point to the astro photography folks as an example.

I think many of us have the tools to construct an image with light, IR, Visible or UV as is described in the paper Keith posted. What we are missing is the software a little hardware to "expose" the stack and the software to "develop" an image. I think all that is needed is a digital camera without a lens and a fast computer to run the software that does the math and tries and make an image from it. No small undertaking. I don't understand the math well enough to write the programs to do it. The math is pretty hairy. But I can see how to go about doing it. Most of the problems that stood in the way were physical limitation of computers in term of the ability to access data fast enough and do the calculations fast enough. With multi core processor computers using multiple hard drives. One core can have the data ready for the one of two cores to do the operations and pass it off to a forth core to write to a second drive with almost no wasted effort and cutting the processing time in half not counting the longer times saved for reading and writing the data to storage.

The way I understand it the data from which the image is calculated from is generated by capturing a lot of data as something non-linear is introduced and scanned across the field. The Electromagnetic (EM) waves can be scanned though a pin hole. Two wave fronts of EM (or Light) can be scanned across the image or many other non-linearities used to generate the data. The image can be moved across the non linear feature, the recoding array moved around to get a different view of the the subject or the illumination scanned over the subject. I expect they all can be combined as well if it can be accounted for in the programing. It just has to be unscrambled with a computer and refined with more computation of the data and more data from more frames.

There is nothing really new or earth shaking about this, it is just finally being able to get computers that are fast enough and cheap enough to devote to running several of them on nothing else for their useful life, but this problem and having software tools developed to the point it doesn't take someone 10 years to do it. If you have four frames of data and it takes a minute to run them it takes 17 hours to run 10,000 with no interaction among the frames. I think the frames interact a lot in this case.

NASA was reconstructing radio signals from Voyager with methods not too different from this since the late 70's. The methods they used to do a 256 byte transform on one dimensional digital data back then are really elegant. They had to be run on the computers then available. It runs about 50 times faster than a FFT, but of course there is no phase data and it only returns digital data, on and off.

When and if someone makes it work I think it will be obvious enough a lot of people can do it. It is not like a piece of hardware that can be clearly patented. There are enough different ways to do it that it would be very hard to tie up as intellectual property. That's why I posted it in the first place. Three or four dimensions 16 bits deep are a lot harder and more complex than one or two dimensions 1 bit deep that NASA did in the 70's but not that much different. You just can't do it on a 1 MHz processor with 16 K of memory and tape drives they used in the 70's.

Gordon

Hi Gordon,
Your comments are more than interesting.
The basic idea of a "lensless light microscope" relying primarily on computing power is really fascinating.

It seems that there are groups in the USA performing experiments very similar to those reported by the Sheffield group. I'll try and get a summary together this weekend. Also, I'll be visiting a lab in Cambridge (MA) later this week and I'll see if I can get some additional input.
In the meantime a Topic summary is now available at:
  http://www.kscitech.com/MicGroup/Msg_16/   (this webpage)

One aspect of such a microscope configuration that we have not yet discussed is the definition of 'magnification'. Also, I suppose, as to how it relates to the effective field of view?

But right now I have to return to some real work ;>(

Regards, Keith

"Gordon Couger" wrote:
> We hardly view the subject directly with a compound microscope...
> ...Almost every other field has gone a lot further with a computer than
> light microscopy. I point to the astro photography folks as an example

Having read the paper again I have come to the following conclusions which may be helpful in summarising the position.

1. The technique draws on standard optical theory, i.e. Abbe et al.

2. Resolution, as in 1 above, is wavelength limited.

3. Although there will not be gains in potential resolution for light microscopy there will be considerable improvements for electron (and shorter wavelength) microscopy because present lens systems are poor.

4. For optical microscopy there is a very simple light source, seemingly not costly to construct, which must emit a single wavelength presumably based on atomic spectral emission.

5. Although, presumably, several "colours" could be used in tandem and combined, a multispectrum image such as produced by brightfield illumination will not be possible. This is unlikely to matter apart from aesthetic considerations.

6. The mechanical part of the optical scope will be very simple. A stepping stage motor.

7. The detector need be nothing more than a reasonably high end device but not a horrendously expensive cooled device.

8. The most costly components will be the processor(s) and memory but these are becoming cheaper.

9. 3D image contruction will become more easy and will be done entirely by software.

10. Working distance to the specimen can be large.

11. Working distance to the detector may be large unless an intermediate lens is used.

12. The maximum resolution and contrast allowed by optical theory may be obtained without recourse to the very expensive objectives (and optical pathway) which currently need to be interchanged for various purposes.

13. It offers little extra to routine light microscopists (e.g. in clinical settings) or amateurs but will be attractive to researchers working near the limits of resolution. However, it could become the routine instrument for many purposes if the overall cost dips below that of conventional microscopes. Yet it is unlikely to wholly replace the need for conventional microscopes since the elucidation of structure is helped by comparison of images produced using differing contrast methods.

14. Fluoresecence microscopy will be supported.

15. There seems no reason in principle why epi illumination should not be supported.

16. The principal intellectual contribution of the authors is in developing algorithms which are greatly faster and more stable than those hitherto and this is aided by their innovative technique of recording multiple overlapping proto-images.

Have I got it wrong or missed a key point?

Selwyn

Hi Keith,
I would think magnification would be some function of the wavelength of the illuminating radiation, the nature of the nonlinearity in size and sharpness, the number and size of steps and the number of passes with the software.

It's not exactly like I haven't been here before. I did a little work on a device with a 1/16 pin hole for gamma rays that measured the activity while scanning a 4 inch rock core and constructing an image as a CAT scan does. If I find software for this method I will try and use that data with it. It is really big steps and only one sensor. It took a week or two to scan a foot of core. It found cracks and pores in rocks that water followed. Sandia used something similar to watch radioactive waste percolate through rock cores.

Gordon

"scitech200" wrote:
> One aspect of such a microscope configuration that we have not yet
> discussed is the definition of 'magnification'. Also, I suppose, as
> to how it relates to the effective field of view?