2.1.

Retinal Pathologies Treatable With an Argon Laser

Three common retinal diseases that are treatable with argon laser therapy are diabetic retinopathy, macular degeneration, and retinal tears.⁷ We provide a brief description of each with an established treatment protocol using argon laser therapy. Keep in mind that all treatments are currently performed manually.

2.2.

Eye Movement

Eye movement is controlled by external ocular muscles: the lateral rectus muscles for looking to the side, the medial rectus muscles for looking toward the nose, the superior rectus muscle for looking up, the inferior rectus muscle for looking down, and the superior and inferior oblique muscles for depressing and elevating the gaze. These muscles act in a coordinated fashion to move the eyes as a conjugate pair. The retinae, as relatively rigid ocular globes, also move in tandem with eye movement as a conjugate pair. The following is a brief review of eye movements listed from fastest to slowest.^{8, 9}

3.1.

Tracking System Development

One of the key improvements to be provided by the computer-assisted system is to guard against undesirable and unanticipated eye movement during laser irradiation. If the eye moves during irradiation, the lesion site will not be where it was planned to be. Not only is it important that the actual lesion location fit into the treatment protocol selected by the ophthalmologist, but a misplaced laser irradiation site could cause severe vision loss. This implies the need to track any eye movements during irradiation and to adjust in real time the pointing of the laser to account for movement in such a way as to maintain the same lesion location. If eye movement exceeds the ability of the system to compensate, then the system must terminate laser irradiation.

Methods have been devised for eye tracking, but most are designed to monitor the anterior eye and cornea. To place lesions on the retinal surface with acceptable accuracy, the system must directly track the retina.¹¹ A few methods of tracking the retina directly have been devised. ^{11, 12, 13, 14} For our purposes here, however, the choice of tracking methods is rather limited. When one accounts for the need to integrate (at reasonable cost) a photocoagulation laser control and pointing subsystem, an automatic lesion pattern generation subsystem, a lesion growth monitoring and control subsystem, and the retinal tracking subsystem of choice, with response times of $5\mathrm{ms}$ —and have it all work together seamlessly—then the most attractive tracking method is some combination of a global digital tracking technique and a high-speed analog technique. This choice is consistent with recommended algorithms for the rigid-body, translation-only tracking problem.¹⁵

3.1.1.

Conceptual design

Initially, the feasibility of a robotic laser system for retinal surgery was studied.⁶ The intent was to “develop an automated laser delivery system and retinal observation system that is capable of placing multiple lesions of predetermined sizes into known locations in the retina.” Markow first demonstrated the concept of the system and identified key research objectives and technical hurdles required to make the system a reality. Over the course of several years and the work of many researchers (as discussed below), the optical design of the overall system evolved to that shown in Fig. 1 . This design, including a digital tracker, an analog tracker, and a confocal reflectometer for lesion control is shown here to help orient the reader for the following discussion.

Fig. 1

Optical design of a computer-assisted laser retinal surgery system, including digital tracker, analog tracker, and confocal reflectometer for lesion control.

3.1.3.

Digital global tracker prototype

Barrett advanced the design of the digital retina tracker.¹⁴ The digital tracking method uses a standard 30 frames per second (fps) charge coupled device (CCD) video camera attached to a nonmydriatic fundus camera to obtain an image of the retina for both tracking and the user interface. The CCD camera is connected to a frame grabber hosted by a standard desktop personal computer (PC). The digital tracking system provides a global retinal tracking capability using a blood vessel template to update the position of the irradiating laser on the retinal surface. Conceptually, a reference image (the tracking template) of the object to be tracked (the retina) is first obtained and a template is generated. When tracking is initiated, the tracking template is swept over incoming images of the object in real time. A match is found between the tracking template and the object image, and object movement information is derived from a change in the location of the template match. If no match can be found, a “loss of lock” signal is activated. Image changes can be categorized as translation, rotation, and/or scale. Translation is object displacement along the image abscissa, ordinate, or both. Rotation is an angular displacement around a specific axis. Scale is an object’s increase or decrease relative to a fixed ratio within an image. The digital tracking method performs very well when compensating for translation. Rotation is not a significant problem, as noted by de Castro : “ $\dots$ if small local deformations are neglected, during fixation the fundus apparently performs rigid plane movements, with small, if any rotational component.”¹⁷ Furthermore, because the distance between the camera image plane and the retina remains unchanged, scale changes are also minimal.

Template matching requires a reference image of the object. For this subsystem, we used six one-dimensional blood vessel templates to assemble a single two-dimensional retinal vessel “fingerprint” template. A pair of horizontal and vertical one-dimensional templates is selected from each third of a retinal area rich in vessels via the graphical user interface (GUI). The horizontal template chosen from the upper third of the reference image is designated as the first horizontal template. The other five one-dimensional template locations are referenced to the first horizontal template’s location. This technique results in a computationally efficient model of the subject’s retinal vessel pattern. See Fig. 2 . With the template defined, the similarity metric to determine the “goodness” of the match between the template and the object image must be defined. There are a number of metrics for this, including maximum absolute error, mean absolute error, and mean-square error. For this design, we used the absolute value of the difference (AVD). The AVD method uses integer math to calculate the absolute value of the current template response minus the expected template response as a measure of match between the template and the object image. This similarity measure proved to be robust and computationally efficient.

Fig. 2

The two-dimensional template used for the digital tracker. (a) Human fundus image with a two-dimensional template superimposed on the image. (b) Diagrammatic representation of a one-dimensional template response in the vicinity of a blood vessel.

There are several different methods of scanning the template over the object image. Ideally, one could calculate the template response at each and every pixel location within the object image, which is called an exhaustive search. While this would almost always yield a match on each object image, it is computationally expensive. For the retinal tracking subsystem, this would require calculating the AVD metric at approximately $60000\mathrm{pixel}$ locations. The frame grabber providing the image data to the PC and a typical PC could not keep up with such a computational load while maintaining a lock on the retina. Another method is to use a coarse-fine technique. In this technique, the template is scanned over the image, regularly skipping over some fixed number of image rows and columns in a coarse search. Based on the template response from the coarse search, a fine (i.e., exhaustive) search is performed locally about the location of the coarse match to find the more exact match location. While this technique speeds up the tracking task tremendously, there is a significant probability that the true match location may be skipped over during the coarse search. With this in mind, we used a limited-exhaustive search technique. In this technique, the object’s velocity characteristics (i.e., the anticipated eye movements) are used to limit the search in subsequent frames to a small subregion of the image. For example, for a camera providing images at a standard rate of $30\mathrm{fps}$ , one can calculate how much the object could theoretically move during the $33\mathrm{ms}$ between frames. Because the location of the object is known in the present frame, we need only search a subregion of the subsequent frame centered about the last known match location. This technique tremendously speeds up the tracking task, yet avoids the danger of skipping over the true match location. However, should the tracked object move faster than anticipated, it will not be present within the subregion on the subsequent frame, and loss of tracking lock will occur.

While the digital tracker worked well, provided excellent integration into the overall system, and allowed the use of an intuitive GUI, its response time did not meet the specifications outlined earlier. While we could have moved to a faster frame rate camera, the higher cost, shorter integration time for the image, and increased data bandwidth would have created more problems than it solved. Thus we investigated a secondary, synergistic tracking method that would provide the necessary speed.

3.1.4.

Analog local tracker

Wright and Ferguson worked together to integrate an analog tracking method into the system to complement the digital tracker.^{18, 19} The analog tracking subsystem locks onto a small area on the retina where the reflectance is significantly different than the surrounding tissue. Often a single laser-induced lesion is used for this, because the higher scattering coefficient of the denatured tissue makes it appear bright compared to the surrounding tissue in fundus images. We move a low-power secondary laser beam in a small circle (“dithering” the beam) around the designated reference lesion (often formed just for tracking purposes at the beginning of the procedure) and detect the returning light with a confocal reflectometer. At any given instant, when this “dithered” beam is pointed on the reference lesion, the reflectometer signal is high; when it is pointed off the lesion, the reflectometer signal is low. Thus the reflectometer output signal varies synchronously (when corrected for phase shifts) with the periodic dither signal which drives the pointing mirrors of the low-power secondary laser.

This output is processed to yield $\mathrm{\Delta }x$ and $\mathrm{\Delta }y$ error correction signals from the confocal reflectometer. The error signals initially increase as the center of the tracking beam’s dither circle moves farther from the center of the reference lesion, and they contain all of the information needed to redirect the tracking beam back to the center of the reference lesion, thereby minimizing the error signal, as illustrated in Fig. 3 . By integrating these error signals over time, cumulative errors produce net displacements. Using these error signals to drive the main steering mirrors in the tracking beam path (which also directs the higher power coagulating beam), arbitrary motions of the reference lesion relative to the tracking beam can be followed. Because the reference lesion is fixed with respect to the retina, this results in retinal tracking. Such a control system is a variation on the servo loop concept.

Fig. 3

Conceptual drawing of the analog tracking method. The $X$ and $Y$ sinusoidal signals sent to the dither beam pointing mirrors are in quadrature, resulting in a circular dither beam path. Phase-sensitive detection of the reflectance data returned from the dither beam, referenced to the phase of the $X$ and $Y$ sinusoidal steering signals, provides $\mathrm{\Delta }x$ and $\mathrm{\Delta }y$ error signals as shown in the diagram.

Figures 3a to 3c depicts some representative dither circle signals and the resulting reflectometer signals. Phase-sensitive detection is accomplished with lock-in amplifiers,²⁰ which are simply analog multipliers (mixers) followed by low-pass filters, which produce dc offset voltages proportional to the amplitude component of the reflectometer signal which is in phase with the associated $X$ or $Y$ dither signal. The resulting pair of error voltages constitute a tracking correction vector that is relayed to the main steering mirrors. Note that in addition to the $\mathrm{\Delta }x$ and $\mathrm{\Delta }y$ error signals provided by the two lock-in amplifiers, a third signal is extracted from the reflectometer output: the average value of the returning light from the dither beam. Figure 3a shows a case in which the dither circle is displaced along the $x$ axis (but not the $y$ axis) relative to the reference lesion. There is a $\mathrm{\Delta }x$ error signal proportional to the displacement and a zero $\mathrm{\Delta }y$ error signal. Note the average value, or dc level, of the confocal reflectometer is significantly greater than zero. Figure 3b represents a case in which the dither circle is centered and locked on the reference lesion (the eye may or may not be stationary at this instant) so that null error signals are produced. Again the dc level of the confocal reflectometer is significantly greater than zero. In Fig. 3c, the dither circle has “fallen off” the reference lesion completely and the error signals are meaningless. But note that the dc level of the confocal reflectometer signal is now very low. This low dc value condition allows such a loss of lock condition to be detected.

While this analog tracker is very fast and meets the response time requirements of the overall system, it is only a local tracker. That is, if the analog tracker “falls off” the reference lesion completely and loses lock, this subsystem has no way to reacquire the tracking lock on its own. However, the global tracking ability of the digital tracker can be brought into play to “help” the analog tracker reestablish the lock. This fact, plus the need for global fundus images used as part of the user interface, continue to make the digital tracker a necessary part of the overall system.

3.2.

Lesion Depth Control Subsystem

As previously mentioned, current argon laser treatment protocols require the physician to preset the laser spot size, power, and irradiation time in hopes of achieving consistent therapeutic lesions across the surface of the retina. This is not a realistic expectation as the absorption properties across the retina are nonhomogeneous. In our investigations, we have seen the same laser parameters provide shallow lesions of nontherapeutic value, “good” therapeutic value lesions, anomalous lesions (with bubbles and other abnormalities), and retinal bleeding on the same retina.^{21, 22} This motivated Jerath to develop the lesion depth control subsystem.^{23, 24} This subsystem’s purpose is to dynamically control laser irradiation time during lesion formation to provide consistent lesions across the retinal surface.

The research goal for this subsystem has been to use visible, measurable lesion reflectance parameters as a noninvasive measure of lesion depth in real time. During therapeutic lesion formation, the reflective properties of the tissue change due to an increase in the tissue scattering coefficient.²⁵ This results in an increase in the amount of diffuse laser light reflected back from the lesion center.

Yang and Jerath used video images of the forming lesions to study lesion reflectance profiles.^{21, 23, 26} Jerath was able to achieve consistent retinal lesions of prescribed depth or diameter using a UNIX-based workstation coupled to a fast frame CCD camera. The system was tested successfully on a model egg white medium and also in vivo on rabbit retinal tissue. Jerath demonstrated the capability to achieve reproducible, nearly identical lesions under conditions of varying absorption and laser power. The camera-based system was able to control lesion parameters in real time; however, the camera field of view required for lesion formation was not compatible with the camera field of view required for retinal tracking. Therefore, two separate cameras would be required using this technique.

To remedy this, Wright and Ferguson investigated the use of a confocal reflectometer to measure diffuse central reflectance of a lesion during laser irradiation. The basic premise behind confocal reflectometry is the use of a pinhole aperture to detect light only from the intended focal plane of interest and reject light from other regions. This allows for the isolation of the desired reflectance signal emanating from the central portion of the forming lesion. Therefore, specular reflection from the cornea or from other reflective surfaces within the system are effectively eliminated. Any reflection traveling along the central beam axis still passes through the pinhole aperture. However, the use of a small opaque beam block in the center of the lens eliminates this noise signal while having little effect on the diffuse reflected beam from the forming lesion.

The magnitude of the diffuse reflected light collected from the forming lesion using a confocal reflectometer has a characteristic, repeatable response, as shown in Fig. 4 . The characteristic curve has three distinct regions: the latency region, the growing region, and the saturation region. The latency region is the period from when the retinal tissue is first illuminated with laser energy until the lesion starts to grow. The latency region is the flat part of the reflectance curve before the curve begins to rise. During this time interval, the retinal tissue absorbs laser light energy but does not achieve a temperature high enough for tissue coagulation. Once sufficient light energy is absorbed and coagulation occurs, the latency region ends and the lesion begins to grow quickly. This results in a corresponding steep rise in the reflectance signal called the growing region. Following this period of rapid growth, the reflectance curve flattens out into the saturation region. Should laser irradiation continue much beyond this point, overexposure will occur and result in significant retinal damage.

Fig. 4

Confocal reflectometry of lesion formation at two different laser power settings. The characteristic latency region, growing region, and saturation region is evident in both curves.

Barrett, de Graaf, and Wright attempted to correlate some measurable parameter of the lesion reflectance signal with the actual lesion depth. If a correlation could be found, the reflectance signal could then be used as a control signal for the laser shutter. They investigated the correlation between the latency period to lesion depth and the slope of the reflectance curve to lesion depth in a retinal phantom. Although these investigations yielded a promising (0.66 to 0.99) correlation coefficient between the visible lesion reflectance parameters and lesion depth, they did not believe the correlation was strong enough to be reliably used to control lesion depth parameters for laser surgery. They then investigated the correlation between the integrated lesion reflectance, the area under the lesion central reflectance curve, and lesion depth. Over the course of extensive experimentation using a variety of laser parameter combinations and purposely injected noise, they found that the correlation coefficient between the integrated central lesion reflectance and lesion depth varied between 0.90 and 0.99, as shown in Fig. 5 . ^{27, 28, 29, 30, 31} Ludwig and Naess developed a microcontroller-based system to capture the reflectance signal from the confocal reflectometer and to produce a laser shutter control signal to develop lesions having consistent parameters.^{32, 33}

Fig. 5

Correlation between the area under the central reflectance curve of laser-induced lesions and the lesion depths.

3.3.

User Interface

One of the advantages the system provides, in addition to greater patient safety, is an easy to use, intuitive GUI developed by Wright. A representative GUI display is shown in Fig. 6 .

Fig. 6

Opening screen of the GUI. The left panel provides status information, the right panel provides menus for selecting various system actions, the bottom panel is used for messages to the user, and the center panel is a real-time video image of the retina provided by the fundus camera. A simulated retinal image is shown here.

Without any computer assistance, it is tedious for the ophthalmologist to manually identify all of the lesion locations required. For treatment of a condition such as diabetic retinopathy, this may be thousands of lesion sites. Each location must be chosen not only for its therapeutic value, but also to avoid critical areas such as the macula and major retinal blood vessels. The computer-assisted system incorporates a single-monitor, mouse-driven human/computer interface that helps the clinician designate the desired lesion locations and perform the surgical procedure.¹⁹

The reader may notice that Fig. 6 is not a ${\mathrm{Microsoft}}^{\circledR }$ ${\mathrm{Windows}}^{\circledR }$ interface. Due to the stringent real-time requirements, low-level real-time operating system software was developed to control the system.

4.1.

Dutch Belted Rabbits

An unretouched fundus image from one of the rabbits after testing is shown in Fig. 7a . In this figure, the tracking lesion used for the analog tracker is the lesion closest to the center of the frame. Two $3\times 4$ patterns of lesions were selected via the system GUI and created during eye motion of up to $32\mathrm{deg}∕\mathrm{s}$ . Each lesion is approximately 250 to $300\mathrm{\mu }\mathrm{m}$ in diameter. The pattern on the left, intended to be rectangular, was created without correction signals from the analog tracker; it reveals the relative movement of the fundus during irradiation and shows why high-speed analog tracking is needed. The irregular lesion placement is due only to eye movement during the irradiation. The lesion pattern on the right was created with the analog tracker in the loop. This pattern shows the benefit of high-speed motion compensation, as well as the accuracy of the lesion pattern that can be created using the digital tracker’s global coordinate system. The actual lesion locations in this pattern were no more than $\pm 2\mathrm{pixels}$ ( $\pm 64\mathrm{\mu }\mathrm{m}$ on the retina) from the intended location specified via the system GUI.

Fig. 7

Photocoagulation on a rabbit fundus, with unlocked and locked tracking, given eye motion of up to $32\mathrm{deg}∕\mathrm{s}$ . (a) Unretouched fundus image. (b) In vivo reflectance signal for a $100\mathrm{ms}$ irradiation on a rabbit fundus during motion-compensated tracking.

In a frame-by-frame analysis of the fundus camera video tape of the entire experiment, the velocities of the rabbit’s reflexive eye movements were determined. The maximum velocity the system could track without losing lock varied according to the optical condition of the cornea and the location of the reference lesion in the fundus field of view (FOV). A clearer cornea permitted faster velocities, as did a reference lesion near the center of the FOV. The lowest velocity at which tracking lock was broken was $30\mathrm{deg}∕\mathrm{s}$ during a short, highly impulsive movement. The highest velocity at which tracking lock was maintained was $38\mathrm{deg}∕\mathrm{s}$ , during a relatively smooth movement. A confocal reflectometer signal of a motion-stabilized lesion forming in $100\mathrm{ms}$ is shown on the right in Fig. 7b.

4.2.

Rhesus Monkeys

The Optical Radiation Branch of the Air Force Research Laboratory’s Human Effectiveness Directorate is charged with performing minimum visible lesion (MVL) threshold studies for a variety of laser conditions. Some technologies, such as the ultrashort pulse laser, do not leave highly visible lesions. This complicates the postirradiation tissue analysis because the precise location of the lesion can be difficult to locate and thus analyze. To remedy this situation, laser professionals at the laboratory place visible row and column marker lesions with an argon laser. Ultrashort laser “lesions” are then placed at the cross section of the visible marker lesions. The visible lesions provide landmarks as an aid in locating the less visible ultrashort lesions in postexposure analysis. The MVL studies are typically accomplished in vivo on rabbits and rhesus monkeys, because the eyes of these species share many tissue properties with the human retina. Although the animal subjects are sedated during these procedures, retinal movement does occur, which complicates the placement of the lesion matrix field.

Barrett and Wright were invited by Optical Radiation Branch scientists to test the digital global tracker prototype system to see if it could be used to place the lesion matrices. An anesthetized primate subject was irradiated with both cw argon laser exposures of $150\mathrm{mW}$ at the cornea for $375\mathrm{ms}$ and single ultrashort pulses of up to $3\mathrm{mJ}$ at $120\mathrm{fs}$ . The cw laser wavelength was $514.5\mathrm{nm}$ and the ultrashort wavelength was $580\mathrm{nm}$ (see Fig. 8 ). Although anesthetized, retinal movement occurred, allowing assessment of the power of the current methodology in precisely tracking the retina. Limited success was achieved due to difficulties in placing the optical system around the primate’s abnormally large nose. Even so, the feasibility of using the system for lesion matrix placement in a primate subject was demonstrated.^{34, 35, 36}

Fig. 8

Retinal image of a rhesus monkey after irradiation with both cw argon laser exposures of $150\mathrm{mW}$ at the cornea for $375\mathrm{ms}$ (lower right) and single ultrashort pulses of up to $3\mathrm{mJ}$ at $120\mathrm{fs}$ (upper left). The CW laser wavelength was $514.5\mathrm{nm}$ and the ultrashort wavelength was $580\mathrm{nm}$ .

5.1.

Retinal Pathology Eye Movement Measurements

Investigations demonstrate that ocular motility in eyes with retinal pathology may show a lower propensity to visit such areas of the retina as compared to nonpathological retinal sites. While current ophthalmic instruments with the ability to image both the retina and a visual function test target placed on the retina have provided this observation, the ability to quantify these images in real time as an eye movement measurement is presently lacking. We adapted the tracking algorithm described above to image the retina using a Rodenstock confocal scanning ophthalmoscope (CSLO) during the performance of a visual fixation task. Direct observation of acuity target placement at the retina under continuous viewing conditions was possible with this apparatus. Target fixation eye movement images of the retina were rapidly digitized from video tape records and registered using the retinal image tracking algorithm. Fixational eye movement pattern densities at the retina were derived from this data.

5.2.

Rat Tracking

We adapted the tracking algorithm to track a rat swimming in a Morris water maze. The Morris water maze was developed by Professor Morris to measure memory retention in rodents. The water maze consists of a 2-m diameter pool filled with an opaque media such as water mixed with powdered milk or dispersed chalk. The opaque media prevents the rodent from obtaining visual cues on the location of a submerged platform. Hooded rats are used because the animals use visual cues to orient themselves, and their vision is more acute than albino rats. The dark fur on a rat’s head provides a highly contrasted object for tracking against the white media, as shown in Fig. 9 , allowing an update on rodent position 15 times per second. The tracking algorithm uses a rotation insensitive tracking template coupled with a limited exhaustive search technique to maintain lock on the moving rat. At the completion of the run, the relative time the rat spent in each quadrant was reported. The tracking algorithm was successfully used to evaluate the time course of memory loss and correlate that to tissue loss in the rat cortex following blunt trauma to the sensory cortex.

Fig. 9

Example tracking results for a rat swimming in a Morris water maze.

5.3.

Computer-Assisted Tattoo Removal

Tattoos are made up of small, membrane enclosed ink granules held within perivascular dermal fibroblasts. Selective laser photothermolysis of tattoo ink is used to disrupt the ink granules without damage to adjacent dermal tissue, thus preventing subsequent scarring. If the skin is targeted by laser energy of a wavelength absorbed by the target ink granule over a brief period of time, the ink granule will be disrupted while cutaneous structures will dissipate the heat to their surroundings without injury. The Food and Drug Administration has approved the Q-switched neodymium:yttrium aluminum garnet $(\mathrm{Nd}:\mathrm{YAG})$ laser for tattoo removal. The laser is set for $10\mathrm{pulses}$ per second with a 3 to $4\mathrm{mm}$ spot size and an energy of 7 to $8\mathrm{J}∕{\mathrm{cm}}^2$ . Typical treatment time to remove a tattoo is four to eight $20\min$ sessions. We believe that the photocoagulation system described above would work well at automated tattoo removal, significantly reducing treatment time. That is, the template matching of the digital tracker could easily be set to follow the tattoo shape without tedious human intervention. Any undesirable movement on the part of the patient would also be automatically compensated for by the system.

5.4.

Blue Light Dermal Imaging

Ultraviolet light (200 to $400\mathrm{nm}$ , sometimes called “Wood’s Light”) has been used since 1903 to image skin disorders. More recently, Faraghan Medical Systems of Philadelphia has used a specialized Polaroid camera to photograph skin disorders with uv light. The uv light is more heavily absorbed in areas containing greater concentrations of melanin.

Overexposure to sunlight can damage the skin. After many years, this damage can cause skin cells to grow at an abnormally fast rate. The lesions formed when this occurs are called actinic keratoses (AK) and solaris keratoses (SK). The lesions formed have a higher concentration of melanin than the surrounding healthy skin. Therefore, when imaged with uv light, the lesions appear much darker than the surrounding skin. Over time, these lesions may grow and multiply forming large, unsightly areas on the skin. Not properly treated, they could lead to even more serious problems such as squamous cell carcinoma.

We are currently doing a preliminary study to use a computer-controlled blue light laser source (Liconix Helium Cadmium laser, $442\mathrm{nm}$ ) to image skin diagnosed with both AK and SK. Melanin strongly absorbs blue wavelengths. Conceptually, a low-power laser will be scanned point-by-point (i.e., a raster scan) across the subject’s face to produce a two-dimensional map of AK and SK locations. The highly adjustable laser pointing and tracking capabilities of the automated laser photocoagulation system is ideally suited to this task and can compensate in real time for patient movement while ensuring safety. The system could even start, stop, and restart the raster scan as needed without compromising the data by using facial landmarks recognized by the tracking system. The patient would then undergo treatment with flourouracil as prescribed by the physician. The treatment success would be monitored with periodic AK and SK mappings. Using the automated system would greatly facilitate the detailed comparison of these mappings with one another to determine treatment progress.