Investigative Ophthalmology & Visual Science Cover Image for Volume 62, Issue 11
August 2021
Volume 62, Issue 11
Open Access
ARVO Imaging in the Eye Conference Abstract  |   August 2021
Millisecond-latency FPGA-based pupil tracker for eye motion stabilization.
Author Affiliations & Notes
  • Bartlomiej Kowalski
    Ophthalmology, Stanford University School of Medicine, Stanford, California, United States
  • Alfredo Dubra
    Ophthalmology, Stanford University School of Medicine, Stanford, California, United States
  • Xiaojing Huang
    Ophthalmology, Stanford University School of Medicine, Stanford, California, United States
  • Footnotes
    Commercial Relationships   Bartlomiej Kowalski, None; Alfredo Dubra, None; Xiaojing Huang, None
  • Footnotes
    Support  NIH grant R01EY031360, R01EY025231, U01EY025477, R01EY028287 and P30EY026877, and Research to Prevent Blindness Challenge Grant.
Investigative Ophthalmology & Visual Science August 2021, Vol.62, 45. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Bartlomiej Kowalski, Alfredo Dubra, Xiaojing Huang; Millisecond-latency FPGA-based pupil tracker for eye motion stabilization.. Invest. Ophthalmol. Vis. Sci. 2021;62(11):45.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : Involuntary eye motion results in image blur or distortion in flood-illumination and scanning ophthalmoscope, respectively. These artifacts are particularly problematic in subjects with nystagmus when using high resolution imaging modalities that require long exposures, including (volumetric) optical coherence tomography and adaptive optics scanning ophthalmoscopy. To address this problem, we developed a low-latency pupil tracker for correction of involuntary eye motion using optical scanners.

Methods : A National Instruments PCIe-1477 Xilinx field-programmable gate array (FPGA) and an I7-6850K Intel central processing unit (CPU) were used in combination for real time estimation of pupil location, size and orientation in images from an acA2000-340kmNIR Basler Camera Link camera. The images, downloaded to the FPGA at 82 million pixels/sec, were formed using a custom Scheimpflug optical setup used off-axis with 940 nm light emitting diode illumination. These features facilitate integration with existing ophthalmoscopes while also avoiding photoreceptor stimulation, thus also being compatible with any vision testing device. A 26 by 19 mm field of view allows imaging of up to an 8 mm diameter pupil while fixation remains within a 40 deg square solid angle.

Results : : A prototype with ~0.4 ms calculation latency when using either 372x400 or 750x744 pixel regions of interest was successfully demonstrated at 510 frames per second. The overall latency is limited by camera readout and download time, (~1-2 ms). Pupil center estimation repeatability with a model pupil was found to be 1-3 um (varies with camera gain). The pupil edge detection algorithm is robust to dark features such as make up and eye lashes, and due to its infrared illumination, it is not affected by skin pigmentation.

Conclusions : An off-axis pupil tracker with sub-millisecond calculation latency and micron repeatability was demonstrated at greater than 500 Hz, for real-time correction of extreme involuntary eye motion. This prototype shows 2-4 times better repeatability than state-of-the-art commercial pupil trackers and lower latency.

This is a 2021 Imaging in the Eye Conference abstract.

 

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×