Automated Setup For Rapid Capture Of Standardized Slit Lamp Examination Image Sets For Use In Virtual Consultation And Ai Aided Diagnostic

Slit lamps are one of the most versatile tools in ophthalmic diagnostics, allowing for the discovery of a vast array of conditions. We want to automate the capture of standardized, high quality slit lamp images in a fast process with low skill requirements for the operator.

This includes:

• images using 3mm slits projected from + and – 45°, covering the whole cornea area
• an image showing the red reflex
• a blue illuminated image for usage with fluorescein
• a white illuminated overview image

Automated slit lamp images could reduce the need for personnel involvement, would facilitate virtual consultation, could be placed in non-medical settings like optician offices and might provide additional insight by enabling AI aided diagnostics.

Two slit projectors, projecting from + and – 45° and a central (0°) camera are mounted to a movable stage.

Each slit projector features an automated light source illuminating an optical slit (using Koehler's principle) followed by a projector lens.

The optical slit can be moved horizontally by a stepper motor, resulting in a sweep of the projected slit across the eye.

Further, a central (0°) spot projector used for the red reflex and a RGBW spotlight for areal illumination are added.

After aligning the device with the patient’s eye the operator triggers the measurement.

An image to evaluate the alignment is recorded, then the scanning sequence is started:

The light source of one slit projector is activated and an image recorded. At the same time the optical slit of the other projector is moved in position for the next image.

This process repeats in an alternating pattern, starting from the apex (with some overlap) and moving outwards, until the whole area of the cornea is covered.

After the scanning sequence is completed the central spot projector is activated and an image of the red reflex is recorded.

Finally, an image using blue illumination and an image using white illumination are recorded.

It is easy to operate the setup and record high quality images. The resolution of the recorded images is currently 1920x1200 pixels with 12 bit color depth.

A total of 40 slit lamp images and 4 additional images (evaluation, red reflex, blue illumination for usage with fluorescein and white illumination) are recorded.

The stepper motors allow fast and precise positioning of the slits so that the whole sequence takes ~ 3 seconds to complete. This makes it easy to avoid eye movement or blinking during the measurement.

The blue and white illuminated images are recorded last, to avoid a narrowed pupil in following images and since they have the highest risk of triggering eye movement or blinking.

Since the performance of the setup on healthy eyes meets the expectations, the next step will be a limited trial with patients depicting selected conditions to further evaluate the abilities of the setup.

If this proves successful further automation of the setup, in the form of an automatic alignment of the device and the eye, will be investigated.

Once enough training data is collected, we want to utilize machine learning methods to aid in the detection of conditions, based on the recorded images.