Purchase this article with an account.
Tobias Rudolph, Stephan Wyder, Sandro De Zanet, Sebastian Wolf, Jens Kowal; A fast auto-focus for digital slit lamp imaging. Invest. Ophthalmol. Vis. Sci. 2013;54(15):1516. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The introduction of digital imaging techniques such as OCT and fundus photography has revolutionized clinical diagnostics. However, the slit lamp biomicroscope remains one of the most widely-used devices, particularly in developing countries. While it is fairly easy to integrate a video camera with a slit lamp, it remains challenging to operate it automatically during examination. Due to the small depth of focus in combination with subtle involuntary movements of the eye as well as the ophthalmologist, the video image gets de-focussed easily. Because the ophthalmologist's accommodation prohibits calibration-based approaches, a fast auto-focus is required.
For this work the slit lamp BQ 900 (Haag-Streit AG, Koeniz, Switzerland) equipped with a 70:30 beam splitter was used. For the focus adjustment, the electrically tunable lens EL-10-30-VIS-LD (Optotune AG, Dietikon, Switzerland) was used. To attach the lens, a custom adapter was build that fits the beam splitter's c-mount connector. In the initial experiments a Prosilica GC1350c camera (Allied Vision Technologies GmbH, Stadtroda, Germany) was used. It captures color images at 1360x1024 pixels and 20 fps, and was connected to the video processing PC via Gigabit Ethernet. To calculate the sharpness of the video image, the gradient-based Tenenbaum measure was implemented. The focus control algorithm continuously measures the sharpness at different lens settings. A bell-shaped curve is then fitted into the control points using least-squares regression. The fitting procedure provides a fast and robust way to find the optimal lens setting. The auto-focus prototype is implemented in C++ and runs on a standard Windows 7 desktop computer.
The system was successfully used to record video sequences of the eyes of volunteers. From a strongly unfocussed setting the algorithm finds the optimal lens settings after 4 frames on average. However, normally the camera is only slightly defocused which leads to much shorter response times of the system.
An auto-focus system is presented that integrates smoothly with existing slit lamp hardware and provides sharp digital images without human interaction. If routinely available it will enable new strategies for patient care, including tele-medical patient monitoring, high-throughput automatic screening, electronic patient records and disease progression analysis.
This PDF is available to Subscribers Only