Automated diagnosis and staging of Fuchs’ endothelial cell corneal dystrophy using deep learning

Abstract

Background To describe the diagnostic performance of a deep learning algorithm in discriminating early-stage Fuchs’ endothelial corneal dystrophy (FECD) without clinically evident corneal edema from healthy and late-stage FECD eyes using high-definition optical coherence tomography (HD-OCT). Methods In this observational case-control study, 104 eyes (53 FECD eyes and 51 healthy controls) received HD-OCT imaging (Envisu R2210, Bioptigen, Buffalo Grove, IL, USA) using a 6 mm radial scan pattern centered on the corneal vertex. FECD was clinically categorized into early (without corneal edema) and late-stage (with corneal edema). A total of 18,720 anterior segment optical coherence tomography (AS-OCT) images (9180 healthy; 5400 early-stage FECD; 4140 late-stage FECD) of 104 eyes (81 patients) were used to develop and validate a deep learning classification network to differentiate early-stage FECD eyes from healthy eyes and those with clinical edema. Using 5-fold cross-validation on the dataset containing 11,340 OCT images (63 eyes), the network was trained with 80% of these images (3420 healthy; 3060 early-stage FECD; 2700 late-stage FECD), then tested with 20% (720 healthy; 720 early-stage FECD; 720 late-stage FECD). Thereafter, a final model was trained with the entire dataset consisting the 11,340 images and validated with a remaining 7380 images of unseen AS-OCT scans of 41 eyes (5040 healthy; 1620 early-stage FECD 720 late-stage FECD). Visualization of learned features was done, and area under curve (AUC), specificity, and sensitivity of the prediction outputs for healthy, early and late-stage FECD were computed. Results The final model achieved an AUC of 0.997 ± 0.005 with 91% sensitivity and 97% specificity in detecting early-FECD; an AUC of 0.974 ± 0.005 with a specificity of 92% and a sensitivity up to 100% in detecting late-stage FECD; and an AUC of 0.998 ± 0.001 with a specificity 98% and a sensitivity of 99% in discriminating healthy corneas from all FECD. Conclusion Deep learning algorithm is an accurate autonomous novel diagnostic tool of FECD with very high sensitivity and specificity that can be used to grade FECD severity with high accuracy.

Global Prevalence of Fuchs Endothelial Corneal Dystrophy (FECD) in Adult Population: A Systematic Review and Meta-Analysis

“…Unfortunately, a reasonable coverage of all world regions was not possible, due to the spurious number of large epidemiological studies on the topic. A vast intercase heterogeneity eventually derived, which is in line with different other systematic reviews and meta-analysis of prevalence [ 15 , 16 ]. Second, we excluded not-in-English publications in this review.…”

“…Nevertheless, the change of prevalence over time is difficult to quantify as it depends on changes of risk exposure and other external factors, such as public awareness of the condition, screening campaign, and diagnostic technological improvements which might in turn modify the clinical approach to the condition. As a fact, the recent implementation of deep learning algorithms has highlighted the potential of these tools in identifying early FECD cases, based on the analysis of one anterior segment-optical coherence scan without additional imaging modalities (e g., pachymetry, specular microscopy, and confocal microscopy) or other information [ 16 ]. The future adoption of such software in clinical practice might in turn determine an increase in the number of people with a diagnosis of FECD, due to the higher sensitivity of our diagnostic toolkit.…”

Digital Image Processing and Development of Machine Learning Models for the Discrimination of Corneal Pathology: An Experimental Model

“…Nonetheless, artificial intelligence (AI) has an immense capacity to learn and analyze a large volume of data and, at the same time, autocorrect and continue learning to improve Photonics 2021, 8, 118 2 of 14 the sensitivity and specificity as a diagnostic and disease progression tool in ophthalmology [3,4]. Recently, supervised ML has been applied to systematic identification and diagnosis of different ocular pathologies, including diabetic retinopathy [5,6], age-related macular degeneration [7][8][9][10], glaucoma [11][12][13], keratoconus [14][15][16][17], corneal edema [18] and Fuchs endothelial corneal dystrophy (FECD) [19], among others. Different deep learning and conventional ML analysis methods are used in ophthalmology; among the most commonly applied are random forest (RF) [20], support vector machine (SVM) [21,22], convolutional neural network (CNN) [23,24] and transfer learning (TL) [25][26][27].…”

Anterior-segment optical coherence tomography for the detection and therapeutic monitoring of corneal disorders

“…In the last couple of years, AS-OCT imaging technology has been the subject for analyzing the microstructural changes of different inflammatory, infectious, degenerative, and dystrophic corneal disorders [5][6][7][8]. Many recent publications demonstrate the tendency to develop machine learning algorithms, efficient enough to help the clinician diagnose and detect disease activity and therapeutic monitoring of various corneal pathologies, including stromal edema associated with angle-closure glaucoma, Fuchs endothelial dystrophy, infectious keratitis, and bullous keratopathy [9,10]. The percentage of diagnostic sensitivity, specificity, and accuracy of artificial intelligence methodologies applied to AS-OCT imaging analysis today has reached 94% to 100%, depending on the study [9][10][11][12][13].…”

“…Many recent publications demonstrate the tendency to develop machine learning algorithms, efficient enough to help the clinician diagnose and detect disease activity and therapeutic monitoring of various corneal pathologies, including stromal edema associated with angle-closure glaucoma, Fuchs endothelial dystrophy, infectious keratitis, and bullous keratopathy [9,10]. The percentage of diagnostic sensitivity, specificity, and accuracy of artificial intelligence methodologies applied to AS-OCT imaging analysis today has reached 94% to 100%, depending on the study [9][10][11][12][13]. However, corneal AS-OCT in its current stage is not exempt from limited capabilities.…”