Manual and automated iris boundary detection provide accurate methodology for ablation centration

Key Points

"Since pupil center may change depending on lighting conditions and the state of accommodation of the patient, iris boundary, which is a stationary reference, is used as the basis for treatment centration, and the cyclotorsional compensation technology used in [this excimer laser] and some other excimer lasers," said Dr. Chernyak, director of research, AMO Development LLC, Milpitas, CA.

"Our study was designed to examine the variability in the physicians' performance and to compare it with the iris boundary detection algorithm of the [company's] laser platform," he continued. "The results show surgeon performance is within limits that would not impact clinical outcomes and that the [excimer laser] image processing algorithm behaves statistically similarly to the ability of expert surgeons in detecting iris boundary. Therefore, it provides accurate treatment centering."

The study included four surgeons identified as experts. They were asked to detect iris boundary in a set of 30 images captured by the laser's onboard, infrared camera. For each image, the surgeons used a software program (AutoCAD, Autodesk) to draw a circle around the iris boundary and note the position of the circle center.

To assess intraindividual reproducibility, i.e., how often the same surgeon chooses the same boundary in a given image, each image was presented three times. All images were presented in random order.

The camera images were captured under real surgical conditions at seven different surgical centers and included images from cases performed using different types of microkeratomes (various mechanical devices and the femtosecond laser) and different refractive procedures (LASIK and various surface ablation techniques).

"The appearance of the images can vary depending on such factors as the hydration state of the eye, illumination, and pupil size, as well as whether or not there is a flap present and what type of instrument was used for flap creation," noted Dr. Chernyak. "About one-third of the images in the study were from procedures with a femtosecond laser-created flap. Good results were still achieved, even though interpretation of those images can be more challenging if there is an opaque bubble layer present or gas breakthrough."

Intraindividual variability

The analysis of intraindividual variability showed an average standard deviation of 53 µm for a given image for a given surgeon. The repeatability among surgeons averaged 76 µm. The standard deviations in the horizontal and vertical directions were identical, which is a favorable finding because it indicates a lack of bias in one direction versus the other, said Dr. Chernyak.

The range of the determined iris boundary center varied from image to image, and the variability was higher in images in which the contrast was poorer.

"Although the standard deviation was less than 100 µm, in some difficult images, the variability was as high as 300 µm, but it was as low as 25 µm for the best-quality images," Dr. Chernyak said. "The more difficult experience with certain images points to the importance of having proper illumination and good focus of the camera."

The answer from the laser's algorithm fell within the physician range for 24 of the 30 images. In the overall analysis, no statistically significant difference existed between the algorithm and manual results.

In perspective

Literature pertaining to optical interpretation of decentration was reviewed to place the results of this study into perspective.

"Our study indicates that using iris boundary for registration should result in treatment decentration less than 100 µm. Literature on this topic suggests that, whether performing a customized or conventional ablation, ablation accuracy within this limit should not induce clinically significant lower- or higher-order aberrations," said Dr. Chernyak.

Further support is derived from a paragraph found in the American National Standards Institute standards relating to laser vision correction systems that establishes a 100-µm limit for lateral resolution and accuracy, he said.

"The findings from our study on surgeon variability and the agreement between surgeon and algorithm performance fit nicely together with the information in the literature about treatment centration limits," Dr. Chernyak concluded.