Self-Interference Fluorescence Microscopy

Comments
Add One

Optical Coherence Tomography (OCT) has a lot of advantages over confocal microscopy, especially for applications where it’s useful to have a large working distance between the probe and the tissue. But a big limitation is that it can only detect reflected light, and so can’t be used with fluorescent stains. Fluorescence is often preferred in conventional microscopy because it allows us to visualise structures that we can’t easily identify in reflectance images. So the race is on to find a way to make OCT work with fluorescent emission as well as reflected light.

In OCT, light from a special kind of optical source is split into two beams. One beam is sent to the sample that we’re trying to image (the eye or skin for example), and the other goes to a mirror. Light reflected back from various layers within the sample is mixed with light reflected from the mirror, and both are sent to a detector. Because the light source has some special coherence properties, the interference pattern at the detector tells us about the depth distribution of reflectors in the sample. If you’re not too familiar with OCT then you might want to have a read of this page to see how this actually works, but the key point here is that all of the useful information (i.e. the image) is contained within the interference pattern.

To be useful, images need to contain some kind of contrast. In OCT, this contrast comes from variations in the amount of light reflected by different layers of the sample. Light and dark regions of the image correspond with high and low reflectivity regions in the sample. This is why OCT produces particularly good images from objects with a well-defined layer structure (like the retina), but not necessarily from more homogeneous samples (such as skin).

In microscopy we often use fluorescent stains to improve contrast. When illuminated with the right excitation wavelength, fluorescent molecules (known as fluorophores) emit light with a second, longer wavelength. This effect is known as the Stokes shift. With the right chemistry, fluorophores can be incorporated into stains which attach to features of interest. If we make a good choice of stains and illumination wavelength, fluorescence microscopy can gives us very high contrast images, showing exactly the features we’are interested in.

Unfortunately, as has been said, fluorescent labels can’t be used with conventional OCT. The reason is simple: the fluorescent emission is totally incoherent with the excitation light. This means that when we mix the emitted light with the reference beam there’s no interference and so no OCT signal and no image. This problem is a pretty fundamental one, but we would desperately like to find a way of performing fluorescence OCT because it would dramatically increase the range of applications for this technology.

Self-Interference

If we can’t interfere the fluorescent emission with a reference beam, the only possibility remaining is to somehow interfere it with itself. The idea of using self-interference to measure the depth position of fluorophores is actually quite an old one1,2, although these techniques relied on a strong reflector lying behind the fluorophore.  But a big step forward came in 2006 with a paper called ‘Fluorescence Coherence Tomography’ from the Wellman Centre at Massachusetts General Hospital (MGH)3.

MGH’s device works by collecting the fluorescent emission from both sides of the sample at the same time. This can be done using a pair of microscope objectives, with one of these also acting as a condenser for the excitation beam. The fluorescence collected by the two objectives is combined and sent to a spectrometer. From this point on the principle is the same as Fourier Domain OCT – fluorescence emission from different depths within the sample results in modulations of varying frequencies in the spectrum. Applying a Fourier transform (after a few corrections) results in a fluorescence-with-depth profile – an ‘A’ scan. A complete cross-section can then be obtained by scanning the sample using a translation tage.

Although this is a clever technique, the images shown in the paper unfortunately weren’t particularly impressive by normal OCT standards. The exact reason for this isn’t clear, but a strong culprit has to be the loss of the optical amplification effect. A conventional OCT signal is essentially the product of the light reflected from a certain depth and the light from the reference arm. Since the reference arm power is typically several orders of magnitude larger than the intensity of light reflected from the sample, this result in a huge, low-noise optical amplification – essentially we improve the signal to noise ratio. This is part of the reason why OCT is sensitive enough to detect the tiny amounts of reflected light which make it back from depths of 1-2 mm in tissue. But in fluorescence coherence tomography one weak fluorescence signal is being mixed with another weak fluorescence signal – so it’s no surprise that the images are much noisier.

Another huge drawback of the approach is that it requires both sides of the sample to be imaged at the same time. This obviously rules out looking at thick samples, as not enough light would make it through to the far side. So tissue would have to be sliced and stained as in conventional histology. One of the biggest advantages of OCT over microscopy is that it can provide high axial resolution with large working distances – making it ideal for use in vivo. Since this device can never be used in vivo for anything other than translucent invertebrates, the advantage over a normal microscope becomes questionable. In any case, it looks as though this line of work has hit a dead-end, as no-one has yet published an improved version.

Phase Shifts

A slightly different approach has been suggested recently4. This paper came from Johannes de Boer’s Group in Amsterdam, which has links with MGH. Unlike the MGH paper, this design only requires that light is collected from one side of the sample. Rather than using a Michelson-type interferometer, the authors put a glass phase mask in the infinity space behind the objective. (Infinity space is where the light is collimated – it’s neither converging to, or diverging from, a focus point).

This ‘phase mask’ is actually just an annulus of glass, so that some of the collected light goes through the hole in the middle, and some goes through the glass ring. The two paths are then recombined with a second lens, and sent to a spectrometer.

The result is that we get a modulation in the spectrum, superficially similar to the modulations observed in Fourier domain OCT. Here, ‘modulation’ means that we see a sequence of peaks and troughs, as though someone has superimposed a sine wave on top of the wavelength spectrum of the fluorescence.

Unlike in Fourier domain OCT, where the frequency of this modulation tells us about the depth distribution of reflectors in the sample, the frequency here doesn’t tell us anything about the depth of the fluorophore. All it tells us is the thickness of the glass in the phase mask, which is something we already know.

The idea is that as the fluorophore moves slightly out of focus, the part of the beam passing through the glass mask accrues an additional delay because the light rays are now at a slight angle to the optical axis. You can see why this happens if you look at the figure below. This delay won’t be sufficient to change the modulation frequency by a detectable amount, but it will be enough to change its phase. By detecting the relative phase change as the point of interest is scanned across the sample, we can build up a depth map.

Figure showing how a change in the position of a fluorophore results in an additional path lenght

Figure 1 : Principle of self-interference fluorescence microscopy. When the fluorophore moves away from the focus, an additional path length is introduced,

One thing to notice is that the whole things only really works if there is a single fluorophore (or cluster of fluorophores) along each ‘A-scan’. If there are multiple fluorophores, then the system will read their average depth, which in most cases is a fairly useless measurement. So the technique is arguably more about profilometry than tomography, and can’t really produce OCT-like images.

Still Searching

Both of these papers have demonstrated that it is possible to do some kind of fluorescence OCT, but with significant limitations. We seem to be a long way from producing fluorescence OCT images which have any kind of clinically significant application.

If you think you know how to do it, suggestions below please!

References

  1. Drabe, K.E., G. Cnossen, and D.A. Wiersma, Localization of spontaneous emission in front of a mirror. Optics comm, 1989. 73: p. 91-95.
  2. Moiseev, L., et al., Spectral self-interference fluorescence microscopy. Journal of Applied Physics, 2004. 96: p. 5311.
  3. Bilenca, a., et al., Fluorescence coherence tomography. Optics express, 2006. 14: p. 7134-43.
  4. de Groot, M., C.L. Evans, and J.F. de Boer, Self-interference fluorescence microscopy: three dimensional fluorescence imaging without depth scanning. Optics express, 2012. 20: p. 15253-62.

Leave a Comment

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Next post: