Eye-Tracker-Guided Non-Mechanical Excimer Laser Assisted Penetrating Keratoplasty

Purpose: The purpose of the study was to implement a new eye tracking mask which could be used to guide the laser beam in automated non-mechanical excimer laser assisted penetrating keratoplasty. Materials and methods: A new trephination mask design with an elevated surface geometry has been proposed with a step formation between conical and flat interfaces. Two recipient masks of 7.5/8.0 mm have been manufactured and tested. The masks have outer diameter of 12.5 mm, step formation at 10.5 mm, and slope of conical surfaces 15°. Its functionality has been tested in different lateral positions and tilts on a planar surface, and pig eye experiments. After successful validation on porcine eyes, new masks have been produced and tested on two patients. Results: The build-in eye tracking software of the MEL 70 was always able to capture the masks. It has been shown that the unwanted pigmentation/pattern induced by the laser pulses on the mask surface does not influence the eye-tracking efficiency. The masks could be tracked within the 18 × 14 mm lateral displacement and up to 12° tilt. Two patient cases are demonstrated. No complications were observed during the surgery, although it needs some attention for aligning the mask horizontally before trephination. Stability of eye tracking masks is emphasized by inducing on purpose movements of the patient head. Conclusion: Eye-tracking-guided penetrating keratoplasty was successfully applied in clinical practice, which enables robust tracking criteria within an extended range. It facilitates the automated trephination procedure of excimer laser-assisted penetrating keratoplasty.


Introduct
Over the elliptical or he laser bea approach ha and specific visual acuity same approa PKP using th both donor a of the graft mask in an a The logic ntegration o movements knowledge, been found amellar ker or laser trep applanation non-contact o the compu   The purpose of the study was to implement a new eye tracking mask which could be recognized by the build-in eye-tracker of the MEL70 excimer laser, and could be used to guide the laser beam in automated non-mechanical excimer laser-assisted PKP.

Materials and Methods
A planar metal ring mask is normally used as a reference object for the eye tracker in MEL70 for refractive surgery. Since in automated laser-assisted PKP a metal mask is also used, it has been decided to leave the camera optics and software intact, and instead, to design a new trephination metal mask which could be detected by the eye tracker software without additional changes in the standard settings of the MEL70 software.
Although eye tracking is not needed for donor trephining, since it is performed on a stationary stage and there are no movements expected, the conventional PKP donor mask (as described in [9]) is suitable for eye tracking without modifications, but since the conventional recipient mask geometry was not suitable for the eye tracking of the MEL 70 excimer laser, a recipient mask has been newly designed to enable eye-tracking for non-mechanical excimer laser-assisted PKP.
The MEL 70 is factory equipped with an active eye tracking unit for refractive surgery which is controlled by image processing software and a built-in PC. It uses a lightweight metal ring mask positioned on the limbus as a reference object. The build-in eye tracker is based on a monochrome 8 bit video CCD camera (the intensity is divided between 0 and 255 gray values) with a resolution of 752 × 582 pixels. It searches for a circular pattern with contrast gradient from light to dark radially towards the image center. It defines a so called "hot zone" where the pupil center needs to be found, unless the laser pulses will be stopped.
Based on the above mentioned concept, two different eye tracking mask designs with different surface geometries have been tested during the development process ( Figure 2). Since the eye tracker operates on detection of a reflected IR image and searching for a circular pattern, an elevated surface design has been proposed with a step formation between conical and flat interfaces. In a reflection image the flat surface appears as bright ring (because of higher reflection), and consequently the conical surface as dark ring, since almost no light gets detected ( Figure 2, arrows).
The masks have been manufactured by the VisioTec company (Adelsdorf, Germany) using stainless steel according to our CAD drawing and specifications ( Figure 3). Two most often used sizes of recipient masks have been manufactured and tested for clinical applicability: 7.5 mm and 8.0 mm, which correspond to 7.6 mm and 8.1 mm donor masks, respectively. Additionally, 6.0 mm masks have also been manufactured and tested for robustness and proof of principle, although this has not been considered for further clinical use. The masks have an outer diameter of 12.5 mm, the step formation was at 10.5 mm in diameter, and the inner diameter is the trephination diameter. The step formation is an interface of flat and conical surfaces with a slope angle of 15°. The thicknesses of the masks differ at the periphery (due to the conical structure of the inner ring) depending on their effective diameters. The orientation notches were replicated exactly as described in the literature [9]: eight triangular shaped orientation notches sized (0.30 mm in base and 0.15 mm in height) and corresponding teeth at the donor mask.

S Sensors 201
The follo

Results
Laser ma data depicte Figure 2(a pigmentation he first mas he eye track pig eye glob  Figure  )) has bee n pattern is sk design ha ker. In cont bs for the pr e 5. Eye tra pulses indu e tracking f n in Figure  oposed      The studi up to 12° til s recommen mask (maxim

Figure image
The eye t observed du before treph

Discussion
There are different implementation strategies for eye tracking [12][13][14][15]. The tracking ensures that the reaction time is essentially shorter than typical movements of the eye during surgery, meaning that the next laser shot will be fired faster than essential movements occur [16]. Various systems are commonly used by different companies: in the previous generation of eye trackers only pupil lateral displacement (2D tracking systems) has been addressed. Nowadays, newly developed systems are equipped with sophisticated tracking features, primarily with closed loop feedback. In recent years, the eye registration has also been introduced to clinical practice [17], being more sophisticated but at the same time technically challenging and very time consuming, since the image processing in a shorter time frame is required: sampling rate is more than 10 times the bandwidth. For the registration usually landmarks on the eye which do not change with lighting conditions, such as the limbus, peripherial iris, or reference marks placed by surgeon have been used [13]. Up to now eye tracking systems in ophthalmology register eye movements in the IRIS plane, hence tracking of the corneal surface was required for our purpose. In contrast to the MEL 70 system, which is equipped with a closed loop eye-tracking system, the open loop systems neglect eye movements during the image capture/processing [18].
Up to now, the alignment of the laser beam generally has been done by the surgeon using an aiming beam (pilot laser), which was positioned onto the center of the mask while the excimer beam is in standby mode. For that continuous monitoring (and corrections if needed) of the laser beam path along the mask interface was necessary. After introducing eye tracking trephination masks, excimer laser-assisted PKP becomes even more efficient: After centering the mask according to limbus and approximately perpendicular to the optical axis (laser beam), the tracking system automatically identifies the edges, positions the beam accordingly, and follows the potential movements.
It has been seen that tilt of the mask from the optical axis of the laser may induce inhomogeneous illumination, resulting in some mismatch of the detected ring from the real edge of the mask. Nevertheless, this mismatch could be corrected via the operating software. It is unlikely to expect laterally or rotationally larger displacements than mentioned above during the surgery.
The conventional recipient mask outer diameter of 13 mm has been used [9], which is larger compared to the new eye tracker masks for the MEL 70, limited by the field of view. The larger mask diameter was initially intended to prevent the sclera from being ablated, since the previous generation of lasers were used with a manual beam manipulation. After establishing the automatic approach of corneal trephination, the trephination mask could have already been optimized, since the MEL 70 yields narrower beam size and supports a precise beam manipulation along the trephination edge.
But even reduction of the mask diameter would not be sufficient, since the laser spots create ablation patterns on the mask surface (see Figure 6(a)), which disturb the homogeneous IR image of the mask surface leading to a incorrect assignment of the tracking landmarks. Moreover, these patterns are changing overtime, so a new mask design with robust tracking criteria was required.
As mentioned earlier, the conventional donor mask was compatible to the build-in eye tracking system due to its round geometry, where the metal surface appears dark in the IR image and provides a changing contrast from bright to dark in its outer edge of the mask. The mask geometry was within the dynamic range of tracking.
The trephination mask geometry has been optimized by introducing a 3D geometry instead of a flat surface in order to achieve stable eye tracking even at a smaller diameter of 10.5 mm. Moreover, the overall diameter has been reduced slightly (only 0.5 mm), to protect the sclera from laser ablation.

Conclusions
This the first time that an eye-tracking-guided penetrating keratoplasty was performed. The use of eye tracking masks represents a useful technical refinement of automated excimer laser-assisted PKP. It even further facilitates the automated trephination procedure. Since the PKP eye tracking using the MEL 70 is performed without modifying the built-in software or any hardware changes, the method could easily be translated into regular clinical routine and may also be implemented easily by other Ophthalmology Departments.