Dr Paul van Saarloos, PhD

1
Effectiveness of Accuracy of Eye Trackers in
Laser Refractive Surgery
American Society of Cataract and Refractive
Surgeons 4-9 April, 2008

• Dr Paul van Saarloos, PhD
• Dr. Mukesh Jain, PhD
• Dr Tarak Pujara, MS

2
CustomVis introduces a new generation of Eye
Tracking

• Dual Limbus based tracking with 1000Hz closed
  loop response
• Gaze tracking.
• Cyclorotation by automatic iris recognition.
• Elevation independent scanning
• (Z data tracked).
• 6 degrees of eye motion tracked!

3
Iris Recognition Cyclorotation automatic
correction
4
Purpose

• Pupil based eye trackers need to image through
  the dry treated corneal surface.
• In addition to that, plume and laser flashes
  could affect the accuracy of such eye tracker.
• The Pulzar Z1, Solid State Laser uses a limbal
  based eye tracker.
• This study is designed to determine if a limbal
  based eye tracker provides an improved accuracy
  in tracking the eye.

5
Eye Tracking?How do we measurePerformance?
6
SpeedandAccuracy
7
Which is Easier to See?Pupil or Limbus?

8
Method

• Analyze video of eye during surgery.
• Manually measure the center of Limbus in each
  frame.
• 3. Compare to eye tracker output for same time
  period and average difference.
• 4. Compare to published measurements of accuracy
  for pupil eye trackers

9
Results
450 consecutive frames were measured. The
average difference between the Eye Tracker output
and Manual measurement was 38 microns

10
Results continued..

• Pupil based eye tracker had an accuracy of
  0.06mm for an intact cornea and 0.1mm for a
  cornea with a flap removed. Accuracy is
  expected to reduce further once the laser
  starts. (Results taken from previous studies)
• While limbal based eye tracker had an
• accuracy of 0.04mm during surgery.

11

• Pupil Tracking Problems
• LASIK Bed Irregular (Eye Tracker can not see
  pupil well).
• Excimer laser dries corneal surface making it
  even less optical.
• Ablation plume and laser flashes also affects
  accuracy.
• Pupil center moves up to 0.7mm as it changes size.

12
Published Data
Very little data exists on eye tracker
Accuracy Talyor et al1 using a special pupil eye
tracker to enhance accuracy, measured errors of
100 microns after a flap was cut (laser was not
fired) Gobbi et al2 also reported 100 micron
accuracy for a pupil based eye tracker, but only
on artificial eyes.

• Determining the accuracy of an eye tracking
  system for laser refractive surgery, N Taylor et
  al, J Refractive Surg. 2000, 16(5)S643-6
• Automatic eye tracker for excimer laser
  photorefractive keratectomy P Gobbi et al, J
  Refractive Surg. 1995 11(3)S337-42

13
To put accuracy measurements in context Bueeler
and Mrochen measured an average induced error of
100 microns when the eye tracker response was
reduced from instantaneous to 1 Hz. (Based on
measured eye movements during surgery and
computer simulations) Pupil based Eye Tracking
creates as much error as a very slow eye
tracker! Simulation of Eye-tracker Latency,
Spot Size, and Ablation Pulse Depth on the
Correction of Higher Order Wavefront Aberration
With Scanning Spot Laser Systems, Bueeler and
Mrochen, J Refractive Surg. 2005 2128-36

14
Summary

• Limbal based Eye Tracking is about twice as
  accurate as Pupil based Eye Tracking

15
Conclusion

• Limbal Based Eye Tracking has the potential to be
  significantly more accurate than Pupil Based Eye
  Tracking during LASIK.
• Limbal Based Eye Tracking allows more accurate
  registration of Custom Treatments.
• Accuracy is as important as speed for eye
  tracking.