Artificial intelligence (AI) and the examination of the cornea at the slit lamp

The slit lamp is an obligatory instrument for every ophthalmologist. It was invented by Alvar Gullstrand in 1910. Over the following decades, the slit lamp has been supplemented by additional applications such as the three-mirror contact lens and the magnifying glass, the applanation tonometer to measure intraocular pressure, various laser applications in post-cataract, glaucoma, diabetic retinopathy and other retinal disorders. The slit lamp has become the most versatile ophthalmic device. However, clinical examination of the cornea has not changed significantly since Gullstrand`s introduction of the slit lamp. The corneal thickness is centrally about half a millimeter or 500 microns. The three major corneal components are the epithelium at about 50 microns, the stroma at 445 microns, and the endothelium at 5 microns. To accurately represent the various changes in the cornea, the direct or indirect cleft slit gap must be precisely focused on the epithelium, stroma and endothelium.  To achieve this, however, very minimal anterior-posterior movements of the joystick are required. This circumstance causes corneal findings to be overlooked.

Technologically, it should be possible to further develop corneal slit lamp diagnostics :  Artificial Intelligence (AI), as an integrated module on the slit lamp, automatically creates images of the corneal epithelium, stroma and endothelium in each patient directly and indirectly in the case of a pharmacologically dilated pupil. The automatic image transfer significantly reduces the risk of overlooking  corneal signs. AI makes it possible to combine several direct and/or indirect image sections into one overall image, ranging from the nasal to the temporal limbus. Thus the history photo-documentation would actually indicate the corneal changes in the sense of a progression or a stoppage. The quality of the photo-documentation of corneal changes would be significantly improved in scientific publications. The current standard of photo-documentation of corneal changes in all ophthalmological journals should be adequately assessed.                                                                                                                                                                           

Additionally a target question and answer game of AI would provide important differential diagnostic findings. A “deep learning” of AI would give the ophthalmologist important hints as to whether or not a corneal alteration is progressive in a given time frame. “Big data bases”, which are also available in an integrated module on the slit lamp would allow individual ophthalmological consultation, for example in keratoconus, cornea guttata and Fuchs endothelial corneal dystrophy, acanthamoeba keratitis and others.      

To be updated shortly..