Feature Of The Week 2/7/10: Manual-Scanning Optical Coherence Tomography (OCT) Probe Based on Position Tracking
Feature Of The Week 2/7/10: Optical Coherence Tomography (OCT) is now an important modality in the field of biomedical imaging for its high resolution and non-invasive nature. People have been developing OCT endoscopes to penetrate into deep biological tissues and extend the limited light penetration depth in near infrared wavelength region. Different ways of steering the probing beam have been developed over the years. As most of them need some form of mechanical actuation, they usually come with the associated cost of additional hardware that must be integrated onto the probes. An OCT probe that can be manually swept over the region of interest by the user and provide a scan of the biological structure along its path, could simplify the imaging procedure. Recently researchers from the California Institute of Technology demonstrated some very interesting work on such an OCT probe. Below is a description provided courtesy of Jian Ren a graduate student in electrical engineering at Caltech.
To achieve the above manual-scanning capability, we designed a new OCT probe based on position tracking. An optical monitoring system was first constructed, where the pose of a simple OCT probe can be continuously tracked while the user sweeps the probe over the region of interest manually. We can reconstruct either planar 2D images or volumetric 3D images, depending on user’s scan pattern, by orienting each OCT depth scan to their individual pose estimated by the system. This method can objectively track and record the actual scan pattern regardless of the scanning mechanism, while other mechanically-actuated OCT probes reconstruct images based on pre-determined scan patterns.
In the tracking system, we employed four infrared LEDs as the feature points. A CMOS camera was used to take their images. We deployed a pair of object detection and tracking algorithms to search the location of the feature points in each frame and find their correspondence among a sequence of frames. A centroid estimator further improved this position estimation accuracy to sub-pixel level. By knowing the accurate distribution of the feature points, we applied a pose estimation algorithm known as Pose from Orthography and Scaling with ITerations (POSIT) to compute the pose. The poses were recorded and related to the concurrent OCT depth scans.
We examined the tracking accuracy of our system by translating the probe in a controlled manner so that we can compare the position estimation with the actual. The results show the system has an accuracy of about 6 microns along two axes and 19 microns along the third. This matches well with the associated imaging resolution of the OCT system.
A phantom sample with carved surface was used to validate this method. First, the probe sitting on a motorized stage scanned the sample’s surface, generating a control image. We then manually scanned the surface with free hands. The images reconstructed by the resolved poses conform much better to the control image than the images reconstructed solely based on their time order. We next imaged a 54-stage Xenopus laevis tadpole using this system. The resulting images exhibit the correction effects of this method and demonstrate its imaging capability.
This new method comes with three major advantages. First, the manufacture of such an OCT imaging probe can be rather easy, as there is no actuation system; the low associated cost makes disposable probes possible. Second, the lightweight design provides the users greater control finesse. Third, it has significant flexibility as it allows the users to manually sweep the region of their interest along any arbitrary path, as long as the trajectory of the device is within the scope of the monitoring system.
For more information see recent Article. Click "Full Screen" for better viewing.