Fig 1.
Optical coherence tomography system design and performance.
(A) Diagram of swept-source anterior segment OCT system. G: galvanometer scanning mirrors, L: lens, RR: retro-reflector, DM: dichroic mirror, BR: balanced receiver. (B) OCT sensitivity fall off performance. Fall off of -6dB was measured at 4.6 mm. The red dashed line denotes -6 dB.
Fig 2.
Schematic control system for pupil tracking ASOCT.
The pupil camera imaged the ocular pupil at 500 frames/second and the acquired frames were used to estimate the eye’s lateral motion. The tracking algorithm calculated the deviation of the segmented ocular pupil from a designated reference position to estimate eye motion. Correction signals that were proportional to the magnitude of the estimated motion were then generated and summed with the OCT scanning waveforms using an analog summing amplifier.
Fig 3.
Optical performance of the anterior segment OCT system with and without tracking.
(A-B) System diagram without tracking (A) and after 2.5 mm of lateral tracking (B). (C-D) Spot diagrams of the extrema of the +/- 15 mm tracking FOV. The airy radius was 39.5 μm as denoted by the black circles and the optical performance was diffraction limited.
Fig 4.
Characterization of pupil tracking system for a 3 mm step response.
(A) Generated FSM driving waveform (input) and response waveform generated by tracking algorithm (output), and the fit to the output step function. (B) Impulse response after calculating the derivative of the step response. (C) Magnitude of the frequency response of the system estimated by calculating the Fourier transform of the impulse response. The -3 dB amplitude was measured at 58 Hz, as denoted by the dashed red line.
Fig 5.
Characterization of motion correction bandwidth.
(A) 2D M-scan acquired during simulated pupil phantom motion without pupil tracking correction. The red box shows the segmented pupil edge. The segmented portion was then smoothed with a Gaussian filter (B) and intensity thresholded (C). Edge detection (D) yielded the pupil motion trace. Motion traces without (E) and with (F) pupil tracking are shown for 1 Hz simulated sinusoidal motion. (G) Motion correction percentage as a function of simulated motion frequency.
Fig 6.
Representative anterior segment OCT motion images acquired from a healthy volunteer without tracking.
(A) Single frame B-scans composed of 1000 A-lines. (B) Registered and averaged (10x) B-scan acquired in repetitive B-scan mode at a frame rate of 100 Hz. (C) Volumetric image composed of 1000 A-lines and 200 B-scans, corresponding to a volume frame rate of 0.5 Hz. Red arrow denotes significant motion artifacts due to patient motion. Scale bars are 1 mm. Yellow arrow denotes an artifact from specular reflection.
Fig 7.
In vivo 2D M-scans acquired from a healthy volunteer with and without tracking.
(A-C) Untracked SVP and motion traces extracted from the 2D M-scan and pupil camera. (D-F) Tracked SVP and motion traces extracted from the 2D M-scan and pupil camera.
Fig 8.
Averaged B-scans generated by summing 125 B-scans acquired at the same location in 0.75 seconds with and without tracking.
(A-B) Averaged, untracked and tracked B-scans after axial registration only. (C-D) Averaged, untracked and tracked B-scans after full (lateral+axial) registration. (E-F) Digitally zoomed images of corneal stroma for untracked and tracked fully registered B-scans. (G-H) Digitally zoomed images of the anterior chamber angle. Scleral striations (red arrow) and Schlemm’s canal (blue arrow) were better resolved with tracking. Scale bars are 1 mm.
Fig 9.
Anterior segment volumetric time series acquired with volumes composed of 500 A-lines/B-scans and 200 B-scans/volume with and without tracking.
(A) Volume corrupted by prominent motion artifacts before activating tracking. (B) Volume during which tracking was activated; image artifact caused by repositioning of the scanners is denoted by the red arrow. (C) Volume acquired with tracking activated. Scale bars are 1 mm.
Fig 10.
Uncorrected residual motion (normalized by subject motion) as a function of latency and motion frequency.
For a latency of 4 ms, our model predicted a -3dB roll-off at 20 Hz for the OCT motion correction bandwidth, which was close to our measured value of 22 Hz.