Witnessing the first sign of retinitis pigmentosa onset in the allegedly normal eye of a case of unilateral RP: a 30-year follow-up.

PURPOSE: A patient initially presented with constricted visual field, attenuated retinal vasculature, pigmentary clumping and reduced ERG in OS only, suggestive of unilateral retinitis pigmentosa (RP). This patient was subsequently seen on eight occasions (over three decades), and, with time, the initially normal eye (OD) gradually showed signs of RP-like degeneration. The purpose of this study was to evaluate which clinical modality (visual field, funduscopy or electroretinography) could have first predicted this fate. METHODS: At each time points, data obtained from our patient were compared to normative data using Z tests. RESULTS: At initial visit, all tests were significantly (p < 0.05) altered in OS and normal in OD. Visual field and retinal vessel diameter in OD reduced gradually to reach statistical significance at the 5th visit and 6th visit (21 and 22 years after the first examination, respectively). In OD, the amplitude of the scotopic and photopic ERGs reduced gradually and was significantly smaller than normal at the 2nd visit (after 11 years) and 3rd visit (after 18 years), respectively. When the photopic ERG was analyzed using the discrete wavelet transform (DWT), we were able to detect a significant change at the 2nd visit (after 11 years) instead of the 3rd visit (18 years). CONCLUSIONS: Our study allowed us to witness the earliest manifestation of an RP disease process. The ERG was the first test to detect significant RP changes. A significantly earlier detection of ERG anomalies was obtained when the DWT was used, demonstrating its advantage for early detection of ERG changes.

State-of-the-art retinal vessel segmentation with minimalistic models.

The segmentation of retinal vasculature from eye fundus images is a fundamental task in retinal image analysis. Over recent years, increasingly complex approaches based on sophisticated Convolutional Neural Network architectures have been pushing performance on well-established benchmark datasets. In this paper, we take a step back and analyze the real need of such complexity. We first compile and review the performance of 20 different techniques on some popular databases, and we demonstrate that a minimalistic version of a standard U-Net with several orders of magnitude less parameters, carefully trained and rigorously evaluated, closely approximates the performance of current best techniques. We then show that a cascaded extension (W-Net) reaches outstanding performance on several popular datasets, still using orders of magnitude less learnable weights than any previously published work. Furthermore, we provide the most comprehensive cross-dataset performance analysis to date, involving up to 10 different databases. Our analysis demonstrates that the retinal vessel segmentation is far from solved when considering test images that differ substantially from the training data, and that this task represents an ideal scenario for the exploration of domain adaptation techniques. In this context, we experiment with a simple self-labeling strategy that enables moderate enhancement of cross-dataset performance, indicating that there is still much room for improvement in this area. Finally, we test our approach on Artery/Vein and vessel segmentation from OCTA imaging problems, where we again achieve results well-aligned with the state-of-the-art, at a fraction of the model complexity available in recent literature. Code to reproduce the results in this paper is released.

Non-uniform Label Smoothing for Diabetic Retinopathy Grading from Retinal Fundus Images with Deep Neural Networks.

Purpose: Introducing a new technique to improve deep learning (DL) models designed for automatic grading of diabetic retinopathy (DR) from retinal fundus images by enhancing predictions' consistency. Methods: A convolutional neural network (CNN) was optimized in three different manners to predict DR grade from eye fundus images. The optimization criteria were (1) the standard cross-entropy (CE) loss; (2) CE supplemented with label smoothing (LS), a regularization approach widely employed in computer vision tasks; and (3) our proposed non-uniform label smoothing (N-ULS), a modification of LS that models the underlying structure of expert annotations. Results: Performance was measured in terms of quadratic-weighted κ score (quad-κ) and average area under the receiver operating curve (AUROC), as well as with suitable metrics for analyzing diagnostic consistency, like weighted precision, recall, and F1 score, or Matthews correlation coefficient. While LS generally harmed the performance of the CNN, N-ULS statistically significantly improved performance with respect to CE in terms quad-κ score (73.17 vs. 77.69, P < 0.025), without any performance decrease in average AUROC. N-ULS achieved this while simultaneously increasing performance for all other analyzed metrics. Conclusions: For extending standard modeling approaches from DR detection to the more complex task of DR grading, it is essential to consider the underlying structure of expert annotations. The approach introduced in this article can be easily implemented in conjunction with deep neural networks to increase their consistency without sacrificing per-class performance. Translational Relevance: A straightforward modification of current standard training practices of CNNs can substantially improve consistency in DR grading, better modeling expert annotations and human variability.

CRB1 heterozygotes with regional retinal dysfunction: implications for genetic testing of leber congenital amaurosis.

PURPOSE: To test human CRB1 heterozygotes for possible clinical or functional retinal changes and to evaluate whether a patient with Leber congenital amaurosis (LCA) with CRB1 mutations not consistent with previously described CRB1 phenotypes carried a modifier allele in another LCA gene. METHODS: Seven unrelated heterozygous carriers of CRB1 mutations underwent phenotyping by full eye examinations (indirect ophthalmoscopy and slit lamp biomicroscopy) and functional testing (standard full-field electroretinography [ERG] and multifocal ERG). For genotyping of the LCA patients and their parents, denaturing high-performance liquid chromatography (dHPLC) analyses were performed, followed by sequence analysis of CRB1, followed by sequence analysis of the AIPL1 and CRX genes to identify a putative modifier effect in a patient with an atypical CRB1 phenotype. RESULTS: Reduced full-field ERG b-wave amplitudes were observed with scotopic -2 dB flash (140 microV; P < 0.05), normal full-field cone ERGs, and significant regional retinal dysfunction on mfERG in five of seven carriers of CRB1 mutations. A known AIPL1 mutation (p. R302L) was identified as a potential modifier allele in a patient with LCA carrying two CRB1 mutations and with a prominent maculopathy. CONCLUSIONS: In human heterozygotes of CRB1 mutations (parents of offspring with LCA), distinctive regional retinal dysfunctions were found by multifocal ERG measurements that were consistent with the focal histologic abnormalities reported for the two CRB1 knockout mice models. This phenotypic finding may identify CRB1 carriers and point to the causal gene defect in affected LCA offspring, significantly facilitating the molecular diagnostic process. Evidence suggests a modifier allele in AIPL1 in a patient with LCA with prominent atrophic macular lesions and homozygous defects in CRB1.

Automatic Screening and Grading of Age-Related Macular Degeneration from Texture Analysis of Fundus Images.

Age-related macular degeneration (AMD) is a disease which causes visual deficiency and irreversible blindness to the elderly. In this paper, an automatic classification method for AMD is proposed to perform robust and reproducible assessments in a telemedicine context. First, a study was carried out to highlight the most relevant features for AMD characterization based on texture, color, and visual context in fundus images. A support vector machine and a random forest were used to classify images according to the different AMD stages following the AREDS protocol and to evaluate the features' relevance. Experiments were conducted on a database of 279 fundus images coming from a telemedicine platform. The results demonstrate that local binary patterns in multiresolution are the most relevant for AMD classification, regardless of the classifier used. Depending on the classification task, our method achieves promising performances with areas under the ROC curve between 0.739 and 0.874 for screening and between 0.469 and 0.685 for grading. Moreover, the proposed automatic AMD classification system is robust with respect to image quality.

Aortic coarctation and the retinal microvasculature.

BACKGROUND: Aortic coarctation has been associated with generalized vascular disease, yet little is known about retinal vascular patterns and their changes over time. OBJECTIVES: The aim of this study is to characterize the nature and extent of retinal vascular disease in adults with aortic coarctation, and explore age-related effects and associations with cardiovascular outcomes. METHODS: A prospective cross-sectional seroepidemiological study was conducted on 60 consecutive adults with repaired aortic coarctation, age 42.4±14.1 years, 61.7% male. In addition to detailed questionnaires, imaging studies, and laboratory testing, high-quality retinal images were acquired by 45° nonmydriatic digital funduscopy. RESULTS: No patient had evidence of hypertensive retinopathy. A distinctive vascular pattern characterized by bilaterally symmetric tortuosity of retinal arteries and veins was observed. Arterial tortuosity was abnormal in 98.3% of patients and decreased with age (P=0.0005). In patients≥45 years, a 1-point increase in the arterial tortuosity score was associated with a 1.5-fold higher risk of cardiovascular complications (i.e., acute coronary syndrome, stroke, cerebral aneurysm, aortic dissection/rupture) [odds ratio 1.50, 95% CI (1.01, 2.24), P=0.0496]. Abnormal venous tortuosity was present in 75.0% of patients and non-significantly correlated with higher levels of serum inflammatory markers (C-reactive protein, fibrinogen, interleukin-6, and tumor necrosis factor-alpha). A higher venous tortuosity score was likewise associated with an increased risk of cardiovascular complications [odds ratio 1.86, 95% CI (1.03, 3.35), P=0.0392]. CONCLUSIONS: Adults with repaired aortic coarctation exhibit a unique retinal vascular pattern characterized by excessive arterial and venous tortuosity that regresses with age. Greater tortuosity is associated with adverse cardiovascular outcomes in patients≥45 years.

Visual evoked potentials and reaction time measurements to motion-reversal luminance- and texture-defined stimuli.

PURPOSE: Previous studies have suggested that compared to first-order (FO) motion stimuli, second-order (SO) motion stimuli required more cortical time to be processed. The purpose of this study was: 1- to verify this claim with Visual Evoked Potential (VEP) and eye-hand Reaction Time (RT) measurements and 2- examine if the VEP and RT responses are similarly modulated by the same trigger features of the stimuli. METHODS: The VEPs and eye-hand RT for motion-reversal luminance- (first-order) and texture-defined (second-order) stimuli were recorded from ten normal human subjects. VEPs and RTs were measured for each motion class at eight different modulation depths (from 3 to 100%). RESULTS: Our results reveal that for stimuli of low contrast levels, the SO-FO timing differences are approximately 100 ms (RT) or 20 ms (VEP), while for contrasts >or= 15-20% (VEP) or >or= 50% (RT), the SO-FO difference is no longer significant (p < 0.007), suggesting either that the brain can no longer distinguish SO from FO stimuli or that in spite of the added complexity of SO stimuli the brain takes equal time to process both. CONCLUSION: Interestingly, the above contrast discrepancy in SO-FO resolution threshold suggests that, compared to the VEP, the more psychophysical RT measurement can process and thus distinguish a larger spectrum of motion stimuli, thus further confirming the latter measure of the retinocortical processing time as a valid alternative to the VEP.