A Low-Cost Eye Tracker for a Refractive Laser Surgery System

Author(s):
Roy E.. Williams PhD - FEO Medical
Jerre M.. Freeman, MD - MECA Surgery Center
Brian Callies - FEO Medical

Industry:
Life Science

Products:
Vision, LabVIEW

The Challenge:
Designing, developing, and installing a low-cost, active, external, add-on system to help keep a patient eye centered within a surgeon-defined refractive laser beam range during refractive laser surgery.

The Solution:
Fabricating an eye-processing system using National Instruments PCI digital I/O and frame grabber hardware, and LabVIEW software.

Accounting for Patient Eye Movement
The excimer laser refractive surgery field has exploded over the past few years with many new vision-correcting laser and algorithm systems. These systems use an ultraviolet excimer laser to change the cornea shape in a calculated pattern, making it possible for the eye to focus properly. For example, in treating myopia, the laser removes tissue from the cornea to flatten it. The laser corrects hyperopia by steepening the cornea, and cures astigmatism by removing tissue in a more complex pattern. The cornea shaping depends on accurate laser beam placement, which can be interrupted or degraded by patient eye movement.

Currently, most FDA-approved broadbeam refractive laser systems and scanning spot systems do not incorporate eye tracking. They require patients to minimize eye movement during surgery by voluntarily fixating their eyes on a small, blinking light located just above their heads. But because the average person makes about five saccadic eye movements (rapid, involuntary movements that are random in amplitude and direction) per second, the laser beam can become decentralized, resulting in degraded laser vision correction predictability and poor visual quality.

Newer-generation refractive laser systems have implemented eye-tracking techniques that move the laser beam internally, via scanning optics, to account for eye movement. This approach provides a higher-accuracy ablation by virtually eliminating shaping error caused by eye movement.

Our system, called the External Active Eye Tracker (EAET), uses National Instruments off-the-shelf products to combine a refractive laser system with an active tracking mechanism. This system monitors the eye and moves the patient (and thus his or her eye), instead of moving the laser beam, without modifying any hardware in the existing refractive laser surgery system.

Highly Precise System Monitors Critical Eye Position
Our PC-based eye-tracking system monitors the patient eye with video captured via a PCI-1409 frame grabber attached to a Sony black and white camera. It illuminates the eye with infrared light at an oblique angle. It also calculates the pupil center exact location and compares it with the original laser beam center location using an image-processing algorithm written in LabVIEW with the IMAQ Advanced Vision library. A PCI-6527 opto-isolated DIO board controls patient chair movement, to adjust patient eye position and counter undesired eye movement.

The eye-tracking algorithm, which uses IMAQ functions callable from LabVIEW, is divided into five sections:

1. Calibrate for presurgery
2. Process the digitized image for pupil recognition
3. Determine the pupil center
4. Calculate the current pupil center offset from the surgeon-defined offset and drive the appropriate surgical chair motors
5. Interrupt the laser fire if the pupil moves outside a certain surgeon-defined range

A highly accurate 4 mm circular target helps daily calibration within ± 2 microns. A technician places the target under the microscope and accurately positions it to the laser system center via the centering reticle. The system energizes the IR illuminator, automatically captures the target via the PCI-1409, calculates the conversion factor from pixels to millimeters, and saves the system center.

Once surgery begins, the algorithm tracks the eye pupil center at a 60 Hz rate. The algorithm energizes the IR illuminator and determines the pupil and pupil center location. To define the contrast between the pupil and the iris, the algorithm applies a transfer function to the image intensity values by creating a bimodal histogram of intensity values. Next, it reverses the intensity values to produce a photometric negative of the image for the binary image processing routines that follow. It then applies a threshold function to set to zero (black) the image intensity values below a predetermined threshold value, while placing the intensity values above the threshold value to all white (a value of 256 in an 8-bit image representation). This makes the pupil totally white against a black background. The next step involves a binary processing technique called dilation to ensure that the pupil is a solid circular object.

The binary image undergoes a characterization process to determine a set of parameters within the image. Since all pupils are nearly the same diameter, we can limit our search of binary objects to a specified range (normally 2-4 mm under refractive laser surgery). By limiting the eye image field-of-view, we are guaranteed to detect only the pupil. After determining the criterion, the system searches for an image matching the specifications and returns them with their area, width, height, and center measurements. We choose an object and use the center for the pupil center, and the width and height to apply a bounding circle around the displayed pupil during the tracking procedure.

Once the system finds a pupil center, it compares the coordinates against the calibrated laser beam center and records the difference. If the difference is larger than the surgeon-defined value (with a ±100 micron default), the system sends a command to the interface module via the PCI-6527, which applies appropriate signals to move the patient chair in the proper direction. But if the difference is smaller than the surgeon-defined value, the system takes no action.

Finally, if the eye center values fall outside the surgeon-defined value (with a ±500 micron default), the system sends a command to the interface module via the PCI-6527, which applies appropriate signals to interrupt the footpedal control and pause the procedure.

Our EAET system also includes a set of diagnostic functions to allow a technician to test the system interface box and video frame grabber card.

Improving Laser Refractive Surgery Outcome
National Instruments hardware and LabVIEW software have proven extremely useful in our external, low-cost, eye tracker system design, development, and implementation. Whereas it would have taken approximately 24 to 30 months to create a similar system using C code and other commercial hardware, using LabVIEW and NI hardware helped our EAET system function as a prototype within approximately 12 months of development.

Browse All Case Studies »