Deep Learning Assisted Epithelial Basement Membrane Dystrophy Detection

Overview

Epithelial basement membrane dystrophy, also known as Map-Dot fingerprint dystrophy or Cogan microcystic dystrophy, is a common bilateral dystrophy of the anterior human cornea. According to one study, it affects approximately 2% of the human population. A more recent study even reported basement membrane changes in 25% of the general population. However, due to its clinical and morphological appearance, the disease is probably often overlooked. Although epithelial basement membrane dystrophy is asymptomatic in many affected patients, there are some important clinical consequences of the disease to consider: Dystrophy is estimated to be the second most common cause of recurrent corneal erosion syndrome and is also an important differential diagnosis of dry eye disease. Therefore, it can cause severe pain in affected patients. In addition, epithelial basement membrane dystrophy plays an important role in the context of cataract surgery, one of the most commonly performed surgeries worldwide: besides the importance of appropriate disease management before surgery to prevent postoperative exacerbation of ocular surface symptoms, epithelial basement membrane dystrophy is also a risk factor for inaccurate preoperative biometry. In recent years, specific features of epithelial basement membrane dystrophy have been introduced in examination methods other than slit-lamp biomicroscopy, such as epithelial thickness mapping or optical coherence tomography. Due to the recent introduction of a variety of deep learning systems, the application of machine learning could significantly increase the detection rate for epithelial basement membrane dystrophy. Furthermore, to the best of our knowledge, the change in disease characteristics over time is currently unknown. Therefore, the first part of this study will investigate the ability of an automated deep learning system using optical coherence tomography scans to distinguish between normal human corneas and corneas affected by epithelial basement membrane dystrophy. For this purpose, 100 eyes of 50 patients will be included in both study groups. In an optional 2nd part of the study, a second visit will be planned in patients with epithelial basement membrane dystrophy to investigate the reproducibility of disease characteristics as a secondary outcome.

This study aims to investigate the capability of an automated deep learning system using anterior segment optical coherence tomography scans to distinguish between normal human corneas and corneas affected by epithelial basement membrane dystrophy. In an optional substudy, a second visit will be scheduled to investigate the reproducibility of disease characteristics as a secondary outcome. One-hundred eyes of 50 patients with epithelial basement membrane dystrophy and 100 eyes of 50 healthy subjects will be included in this study. After successful screening, all study participants will undergo one single study visit. During this visit, two questionnaires (Ocular Surface Disease Index, Quality of Vision), two different anterior segment optical coherence tomography devices (MS-39, Anterion), a slit lamp examination including slit lamp photography will be performed. In an optional substudy, patients with epithelial basement membrane dystrophy will have a second visit, to compare the variability of disease characteristics, including number of maps, dots, fingerprint lines and cysts between the two visits.