Endomicroscopy Through a Multimode Fibre

Add One

Most modern endoscopes use miniaturised cameras to capture macroscopic images, but this technology is not suitable for very high resolution endomicroscopy systems. The current generation of endomicroscopes either have a scanning head at the distal tip of the probe, or use a fibre bundle to relay the image out of the patient. The fibre bundle approach allows for the smallest diameter probes, but also has disadvantages, including a severe resolution/field of view trade-off. A recent paper in Physics Review Letters has suggested there might be another possibility. The authors managed to transmit an image through a single multimode optical fibre with a diameter of only 200 microns. They achieved a resolution of around 2 microns and a field of view equal to the fibre core diameter, opening up the prospect of an ultra-thin endomicroscope reaching parts of the body which are currently inaccessible.

The general idea of this new endomicroscope is to use the different modes of a multimode optical fibre as ‘pixels’, in a way analogous to how individual cores in a fibre bundles are used.  For an introduction to the idea of modes in optical fibres take you can take look at this page, but for the purposes of this post we can just think of them as different ‘paths’ that light can take through the fibre. There are a finite number of these modes or ‘paths’ which the fibre can support, a number which depends on the fibre diameter. This has a practical – although not physical – similarity to how the diameter of a fibre bundle limits the number of cores. However, unlike in a fibre bundle where each core transfers information from a specific spatial location, the signal from a single point would instead be encoded across a large number of different modes.

It turns outs that the ‘modes per unit area’ of a multimode fibre is around 50-100 times larger than the ‘cores per unit area’ of a fibre bundle. So – in principle, at least – it should be possible  to transmit an image with up to 10 times better resolution (in each direction) for a given fibre size. The problem lies in finding some way to ‘encode’ the image in these modes at one end, and then ‘decode’ it at the other.


Several groups have tried to transmit an image through a multimode fibre using various means. A paper published in 20121 showed that a focused spot can be created at the far end of a fibre using a method called digital phase conjugation. The idea is to first perform a calibration by focusing a spot of laser light at one end of the fibre. A speckle pattern is formed at the far end because of interference between light travelling through different modes. Holography techniques can be used to reconstruct a 2D map of the phase of the wavefront exiting the fibre.

A spatial light modulator is then used to create same phase pattern – i.e. the same wavefront – and send it back through the fibre in the opposite direction. The result is a focused spot of light back at the original fibre end. The effect of the fibre has essentially been ‘reversed’. What’s more, if the focus position of the calibration input beam is moved across the fibre, then it’s possible to measure the exit wavefront corresponding to each position. Then, by using the spatial light modulator to generate each of these wavefronts in sequence, the spot can be scanned through the different positions.

This is all very clever, but unfortunately doesn’t allow us to image through the fibre. The authors of the Physics Review Letters paper have found a way to modify this idea to obtain an image2. They began by scanning a beam over the ‘patient’, or distal, end of the fibre. At the other end -the ‘proximal end’ – they then recorded holograms corresponding to each scan position. This created a transmission matrix, describing how information from different points at the patient end is encoded in the wavefront at the distal end.

When an image is obtained, light comes from not from a single point, but from many points. The wavefront at the distal end is then a super-position (a linear combination) of the wavefronts from each point, with the strength of each contribution depending on the intensity of light reflected from the corresponding point. By inverting the transmission matrix they had previously measured, the authors were able to transform the output wavefront back into an image.

So this allows an image to get from one end of the fibre (the patient end) to the other. Unfortunately, this isn’t all that’s needed: the tissue first has to be illuminated. We can do this, as you might expect, by sending light down the fibre. The problem is that this light has to be coherent – from a laser – in order for the holograpy to work. And when we send coherent light down a multimode fibre, we obtain a speckle pattern at the far end. So the resulting images are highly speckled, and essentially unusable.

The authors’ solution to this was fairly simple. They scanned the illumination beam over the proximal end of the fibre, resulting in different combinations of fibre modes being coupled at different times. The result was a constantly changing speckle pattern at the distal end. They then reconstructed images for 50 of these different speckle patterns and took an average. Thanks to some convenient properties of speckle (essentially that the time-average of a varying speckle pattern is the ‘true’ intensity) they obtained a high quality image.

Things get even better. Because the authors have access to the phase of the returning light, they can perform a kind of digital refocusing – obtaining images from different depths within the sample simultaneously. So the frame rates of 1 Hz which was quoted in the paper is actually for the acquisition, although not the processing, of an entire 3D cube.


The authors are very positive about the potential impact of their invention, going as far as to claim that it may “revolutionize endoscopy in various fields encompassing medicine and industry.” Unfortunately, while the results reported in this paper are impressive as a proof-of-concept exercise, the method is nowhere near being ready for clinical use. Some problems, like the low frame rate, are purely engineering ones; the authors suggest a faster camera is all that is required. Similarly, the long processing times could presumably be overcome with dedicated hardware.  But even once these problems are solved, some pretty big limitations remain.  Firstly, since the method relies on interferometry, it can’t be used with fluorescent agents. Secondly, the lack of a lens means that the imaging plane must be very close to the tip of the fibre, even with digital refocusing. This may not be optimal for many applications.

By the far the biggest problem is a truly fundamental one: the transmission matrix calibration is only valid providing that the fibre doesn’t bend! Actually, the authors show that they can still obtain images using the original calibration when the tip is translated by a small distance, which is actually a fairly impressive achievement. But it avoids the real problem – if you want to put this inside a patient then it’s going to bend a lot. And as soon as it does, the calibration becomes invalid and the probe no longer works. What is needed is a way to recalibrate ‘on-the-fly’ using some kind of reference signal, something like how astronomers correct for atmospheric turbulence based on the image of an artificial ‘guide-star’. At least one group is working on just that, although it remains to be seen how well – and how fast – this can be made to work.

The authors acknowledge this problem, and suggest that the device could be used as a rigid needle probe, avoiding any bending. This does seem to be the most promising application, at least in the medium term. While very thin needle probes can be fabricated using GRIN lenses (see this recent paper from David Sampson’s Group), there is no way of building a GRIN needle probe that could match the field of view of this new method. And it certainly couldn’t be done using fibre bundles. Add in the possibility of 3D imaging, and there is potential for a real breakthrough in minimally-invasive microscopic imaging.


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>