Motion Compensated Handheld OCT

Comments
Add One

I recently discussed several ways of generating a simple optical coherence tomography (OCT) cross-section by manually scanning a fibre optic probe. The idea is that we simplify the engineering of a miniaturised OCT system by dispensing with the need to incorporate scanning mirrors.  It looks like this is a popular topic at the moment because another paper on the theme of handheld probes has just appeared in Biomedical Optics Express. This article takes things a little further, and tries to use motion compensation to improve the stability of the image. This unfortunately re-introduces some complexity into the system, but the authors think it could have still have applications for image guided surgical interventions.

Shaky Hands

Whenever we want to image human beings at a high resolution we inevitably run into the problem of their motion. Breathing and other involuntary movements tend to introduce artefacts into the image and make it difficult to maintain a constant view. If we want to use hand-held probes then things becomes even worse, because we now have to worry not only about the motion of the patient, but the shaking of the operator’s hand as well!

In many ways, OCT is an ideal technique for avoiding motion artefacts. Modern OCT systems can acquire images very quickly, often at a rate of tens or even hundreds of frames per second. This tends to reduce the artefacts within an individual image quite significantly, even for handheld probes. Unfortunately, if we want to perform manual scanning then the frame rate necessarily drops drastically – it may take several second to capture even a single image. On this kind of time-scale there is likely to be significant motion and the image will be distorted.

Luckily, OCT has another important characteristic we can use to our advantage. As long as we can pick out the top surface of the sample then we can measure the distance between the probe and the tissue. So, in principle, we can correct for any motion in the axial direction, either using software or by physically re-positioning the probe. The group from John Hopkins University have done a little of both.

Motion Compensation

The paper’s1 authors built a device containing a piezo-electric motor which was able to re-position the imaging fibre at high speed. The distance of the fibre from the tissue surface was calculated in real time from OCT A-scans using a peak detection algorithm. A feedback control scheme then controlled the motor using this positional information, with the aim of providing a constant separation between the fibre and the surface. The motor was updated at a rate of 460 Hz, which is faster than most of the frequency components of hand tremor.

Images shown in the paper demonstrate that this method successfully removes a great deal, but not all, of the hand tremor. Unfortunately, the device doesn’t know whether an apparent change in the probe-to-surface distance has resulted from hand motion or is a real physical feature of the surface. So, when the probe is scanned across the tissue, true variations in the height of the surface will also be removed from the image.

The authors recognised this problem and tried to re-create the true surface variations. Their approach was to calculate cross-correlations between adjacent A-scans. If there has been an axial shift, the cross-correlation should tell us the magnitude of that shift. This can then be used to adjust the positions of the A-scans in order reconstruct the true surface shape. The examples shown in the paper seem to indicate that this method works to some extent, although the reconstruction is not perfect by any means. Used on its own, without probe re-positioning, this approach will also correct much of the distortion due to hand motion, although it certainly doesn’t do as well as the motorised motion compensator.

Cross-correlations were also used to estimate the true lateral position of each A-scan. This idea isn’t new and I’ve discussed it before, so I won’t repeat it all again. It appears to work fairly well in this case, but suffers from all the same disadvantages I described previously.

Impact

The real impact of this paper is difficult to assess. Certainly the motion corrected images look much better than the originals. However, the real question is whether or not they are good enough to make manual scanning a viable technique. If we have to attach a vertical positioner to the probe then we have already severely restricted what we can do. We couldn’t make a flexible endoscopic OCT probe in this way, for example. In which case it’s arguably not much more effort to add a miniature scanning mirror and obtain high frame rates and artefact free images without the difficulties of manual scanning.

References

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>