Quality Control in the Corneal Bank with Artificial Intelligence: Comparison of a New Deep Learning-based Approach with Conventional Endothelial Cell Counting by the "Rhine-Tec Endothelial Analysis System". / Qualitätskontrolle in der Hornhautbank mit künstlicher Intelligenz: Vergleich eines neuen Deep-Learning-basierten Ansatzes mit der konventionellen Endothelzelldichtenbestimmung durch das "Rhine-Tec Endothelial Analysis System".

Endothelial cell density (ECD) is a crucial parameter for the release of corneal grafts for transplantation. The Lions Eye Bank of Baden-Württemberg uses the "Rhine-Tec Endothelial Analysis System" for ECD quantification, which is based on a fixed counting frame method considering only a small sample of 15 to 40 endothelial cells. The measurement result therefore depends on the frame placement and manual correction of the cells counted within the frame. To increase the sample size and create higher objectivity, we developed a new method based on "deep learning" that automatically detects all visible endothelial cells in the image. This study aims to compare this new method with the conventional Rhine-Tec system. 9375 archived phase-contrast microscopic images of consecutive grafts from the Lions Eye Bank were evaluated with the deep learning method and compared with the corresponding archived analyses of the Rhine-Tec system. Means, Bland-Altman and correlation analyses were compared. Comparable results were obtained for both methods. The mean difference between the Rhine-Tec system and the deep learning method was only - 23 cells/mm2 (95% confidence interval - 29 to - 17). There was a statistically significant positive correlation between the two methods, with a correlation coefficient of 0.748. What was striking in the Bland-Altman analysis were clustered deviations in the cell density range between 2000 and 2500 cells/mm2 - with higher values in the Rhine-Tec system. The comparable results for cell density measurement values underline the validity of the deep learning-based method. The deviations around the formal threshold for graft release of 2000 cells/mm2 are most likely explained by the higher objectivity of the deep learning method and the fact that measurement frames and manual corrections were specifically selected to reach the formal threshold of 2000 cells/mm2 when the full area endothelial quality was good. This full area assessment of the graft endothelium cannot currently be replaced by deep learning methods and remains the most important basis for graft release for keratoplasty.

NEGATIVE VESSEL REMODELING IN STARGARDT DISEASE QUANTIFIED WITH VOLUME-RENDERED OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY.

PURPOSE: To quantify retinal vasculature changes in Stargardt disease1 (STGD1) with volume-rendered optical coherence tomography angiography. METHODS: Optical coherence tomography angiography volumes from healthy subjects and two subgroups of patients with STGD1 with the presence/absence of definitely decreased autofluorescence areas were compared. Optical coherence tomography angiography vessel surface area and vessel volume were measured in central zones (Z) of 1-, 2-, and 3-mm diameter. RESULTS: Twenty nine eyes of 15 patients with STGD1 (20/9 eyes with/without definitely decreased autofluorescence) and 30 eyes of 15 controls contributed data. An enlarged foveal avascular zone was found in patients with STGD1 without and even more with definitely decreased autofluorescence associated with a vessel rarefication in central and also paracentral zones with unnoticeable autofluorescence. Vessel surface area and vessel volume were reduced in both STGD1 subgroups for all zones (P < 0.0001). Stargardt disease 1 eyes when compared to without definitely decreased autofluorescence showed reduced vessel surface area and vessel volume in Z2+3 (both P < 0.05). CONCLUSION: Volume rendering of optical coherence tomography angiography in STGD1 shows a reduced retinal flow in the central macula. This is most likely secondary to loss of neurosensory tissue with disease progression and therefore not likely be favorably influenced by gene transfer and retinal pigment epithelial transplantation. Retinal blood flow assessed by 3D volume-rendered optical coherence tomography angiography could serve as surrogate marker for vascular changes of the central retina.

ZEISS PLEX Elite 9000 Widefield Optical Coherence Tomography Angiography as Screening Method for Early Detection of Retinal Hemangioblastomas in von Hippel-Lindau Disease.

Purpose: To evaluate early detection of retinal hemangioblastomas (RHs) in von Hippel-Lindau disease (VHLD) with widefield optical coherence tomography angiography (wOCTA) compared to the standard of care in ophthalmologic VHLD screening in a routine clinical setting. Methods: We conducted prospective comparisons of three screening methods: wOCTA, standard ophthalmoscopy, and fluorescein angiography (FA), which was performed only in uncertain cases. The numbers of detected RHs were compared among the three screening methods. The underlying causes for the lack of detection were investigated. Results: In 91 eyes (48 patients), 67 RHs were observed (mean, 0.74 ± 1.59 RH per eye). FA was performed in eight eyes. Ophthalmoscopy overlooked 25 of the 35 RHs detected by wOCTA (71.4%) due to the background color of the choroid (n = 5), small tumor size (n = 13), masking by a bright fundus reflex (n = 2), and masking by surrounding retinal scars (n = 5). However, wOCTA missed 29 RHs due to peripheral location (43.3%). The overall detection rates were up to 37% on the basis of ophthalmoscopy alone, up to 52% for wOCTA, and 89% for FA. Within the retinal area covered by wOCTA, the detection rates were up to 46.7% for ophthalmoscopy alone, up to 92.1% for wOCTA, and 73.3% for FA. Conclusions: The overall low detection rate of RHs using wOCTA is almost exclusively caused by its inability to visualize the entire peripheral retina. Therefore, in unclear cases, FA is necessary after ophthalmoscopy. Translational Relevance: Within the imageable retinal area, wOCTA shows a high detection rate of RHs and therefore may be suitable to improve screening for RHs in VHLD.

Optic Nerve Head Volumetry by Optical Coherence Tomography in Papilledema Related to Idiopathic Intracranial Hypertension.

Purpose: Idiopathic intracranial hypertension (IIH) leads to optic nerve head swelling and optic atrophy if left untreated. We wanted to assess an easy to perform volumetric algorithm to detect and quantify papilledema in comparison to retinal nerve fiber layer (RNFL) analysis using optical coherence tomography (OCT). Methods: Participants with and without IIH underwent visual acuity testing at different contrast levels and static perimetry. Spectralis-OCT measurements comprised standard imaging of the peripapillary RNFL and macular ganglion cell layer (GCL). The optic nerve head volume (ONHV) was determined using the standard segmentation software and the 3.45 mm early treatment diabetic retinopathy study (ETDRS) grid, necessitating manual correction within Bruch membrane opening. Three neuro-ophthalmologists graded fundus images according to the Frisén scale. A mixed linear model (MLM) was used to determine differences between study groups. Sensitivity and specificity was evaluated using the area under the receiver-operating characteristic (ROC). Results: Twenty-one patients with IIH had an increased ONHV of 6.46 ± 2.36 mm3 as compared to 25 controls with 3.20 ± 0.25 mm3 (P < 0.001). The ONHV cutoff distinguishing IIH from controls was 3.97 mm3 (i.e. no patient with IIH had an ONHV below and no healthy individual above this value). The area under the curve (AUC) for ONHV was 0.99 and for the RNFL at 3.5 mm 0.90. The Frisén scale grading correlated higher with the ONHV (r = 0.90) than with the RNFL thickness (r = 0.68). ONHV measurements were highly reproducible in both groups (coefficient of variation <0.01%). Conclusions: OCT-based volumetry of the optic nerve head discriminates very accurately between individuals with and without IIH. It may serve as a useful adjunct to the rating with the subjective and ordinal Frisén scale. Translational Relevance: A simple OCT protocol run on the proprietary software of a commercial OCT device can reliably discriminate between normal optic nerve heads or pseudo-papilledema and true papilledema while being highly reproducible. Our normative data and OCT preset may be used in further clinical studies.