Purchase this article with an account.
Pablo Ortiz, Mark Draelos, Ryan P McNabb, Ruobing Qian, Christian Viehland, Anthony N Kuo, Joseph A Izatt; Robotically Aligned OCT for Subject Eye Translation and Rotation Compensation. Invest. Ophthalmol. Vis. Sci. 2020;61(7):2575.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Optical Coherence Tomography (OCT) systems used in the clinic require patient stabilization and gaze fixation to reduce motion artifacts and align the scanner with the region of interest in the retina. This limits the use of OCT for uncooperative patients, such as unconscious patients undergoing surgery. By tracking eye location and gaze, and mounting a retinal OCT scanner to a robot, we can automatically align the scanner with the subject’s eye. We demonstrate the capabilities of our system with retinal OCT images of freestanding subjects.
For retinal scanning, we utilized a 100kHz swept-source OCT engine. The system’s scanner (Fig. 1), which was designed for a 16-degree field of view, was mounted on a 6 degree-of-freedom robot (UR3, Universal Robots) along with three cameras and an IR LED illumination ring for high-resolution pupil and gaze tracking. The IR LEDs illuminated the pupil and produced corneal reflections that were captured in real-time by the cameras. For pupil tracking, the centroid of the pupil's image at each camera was segmented and, through linear triangulation, used to calculate the pupil center in space. For gaze tracking, the images of the corneal reflections were segmented in real-time and used to calculate the optical axis of the eye. This axis was used to adjust the robot orientation during the imaging session such that the mounted scanner aligned with the region of interest at the retina. We tested tracking performance using a mechanically translated and rotated eye model, and demonstrated in vivo operation by tracking and imaging the optic nerve and fovea of a freestanding a human subject without a fixation target (Fig. 2) consented under an IRB-approved protocol.
Gaze tracking tests demonstrate 0.337° accuracy and 0.236° precision within a 28° range. Pupil tracking tests demonstrate 19.5μm lateral and 34.2μm axial accuracy along with 5.9μm lateral and 6.1μm axial precision within a 300mm lateral and 150mm axial range. Human imaging demonstrates stabilization of images automatically tracked to subject retinal features.
We demonstrated OCT imaging (Fig. 2) in freestanding subjects with arbitrary gaze angles.
This is a 2020 ARVO Annual Meeting abstract.
a) OCT scanner mounted on a robot, b) cross section of the scanner design
En face projections of OCT volumes (a and b) and B-scans (c and d) automatically tracked to a selected eye of a freestanding subject's optic nerve head and fovea, respectively.
This PDF is available to Subscribers Only