An AI model to estimate visual acuity based solely on cross-sectional OCT imaging of various diseases.

PURPOSE: To develop an artificial intelligence (AI) model for estimating best-corrected visual acuity (BCVA) using horizontal and vertical optical coherence tomography (OCT) scans of various retinal diseases and examine factors associated with its accuracy. METHODS: OCT images and associated BCVA measurements from 2,700 OCT images (accrued from 2004 to 2018 with an Atlantis, Triton; Topcon, Tokyo, Japan) of 756 eyes of 469 patients and their BCVA were retrospectively analysed. For each eye, one horizontal and one vertical OCT scan in cross-line mode were used. The GoogLeNet architecture was implemented. The coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE) were computed to evaluate the performance of the trained network. RESULTS: R2, RMSE, and MAE were 0.512, 0.350, and 0.321, respectively. R2 was higher in phakic eyes than in pseudophakic eyes. Multivariable regression analysis showed that a higher R2 was significantly associated with better BCVA (p < 0.001) and a higher standard deviation of BCVA (p < 0.001). However, the performance was worse in an external validation, with R2 of 0.19. R2 values for retinal vein occlusion and age-related macular degeneration were 0.961 and 0.373 in the internal validation but 0.20 and 0.22 in the external validation. CONCLUSION: Although underspecification appears to be a fundamental problem to be addressed in AI models for predicting visual acuity, the present results suggest that AI models might have potential for estimating BCVA from OCT in AMD and RVO. Further research is needed to improve the utility of BCVA estimation for these diseases.

Cytokine profiles of macular neovascularization in the elderly based on a classification from a pachychoroid/drusen perspective.

PURPOSE: To classify macular neovascularization (MNV) based on pachychoroid and drusen features and to examine the aqueous humor cytokine signatures of each group. METHODS: In total, 106 consecutive eyes with treatment-naïve MNV and 104 control eyes were examined. The aqueous humor concentrations of 15 cytokines were compared among the MNV groups classified based on the presence of drusen and/or pachychoroid features. Multidimensional scaling analysis was used to visualize the similarity level of the MNV subtypes according to their cytokine profiles. RESULTS: Thirty-one, 18, 43, and 10 eyes were classified into the pachychoroid-associated, drusen-associated, pachychoroid/drusen-associated, and non-drusen/non-pachychoroid MNV groups, respectively. Compared with the control group, cytokines were differently upregulated among the MNV groups. CRP and CXCL12 were significantly upregulated in all MNV groups, whereas CXCL13 and IL-8 were significantly upregulated in three MNV groups, excluding the non-pachychoroid/non-drusen-associated MNV group. Ang-2 was significantly upregulated in three MNV groups except the drusen-associated MNV group. PlGF was significantly upregulated in the pachychoroid-associated and drusen-associated MNV groups. CCL-2 was significantly upregulated in the pachychoroid-associated and pachychoroid/drusen-associated MNV groups. VEGF was downregulated in the pachychoroid-associated and drusen-associated MNV groups, respectively. Multidimensional scaling analysis showed a distinct cytokine profile for each MNV group. CONCLUSION: All MNV groups showed distinct cytokine profiles. Eyes with "neovascular age-related macular degeneration with drusen and concomitant pachychoroid" may share a similar etiology to those with "pachychoroid neovasculopathy" and "choroidal neovascularization with drusen," but have a distinct etiology to those without these. These findings suggest the importance of evaluating drusen and the choroid during the diagnosis of neovascular age-related macular degeneration.

Predicting central choroidal thickness from colour fundus photographs using deep learning.

The estimation of central choroidal thickness from colour fundus images can improve disease detection. We developed a deep learning method to estimate central choroidal thickness from colour fundus images at a single institution, using independent datasets from other institutions for validation. A total of 2,548 images from patients who underwent same-day optical coherence tomography examination and colour fundus imaging at the outpatient clinic of Jichi Medical University Hospital were retrospectively analysed. For validation, 393 images from three institutions were used. Patients with signs of subretinal haemorrhage, central serous detachment, retinal pigment epithelial detachment, and/or macular oedema were excluded. All other fundus photographs with a visible pigment epithelium were included. The main outcome measure was the standard deviation of 10-fold cross-validation. Validation was performed using the original algorithm and the algorithm after learning based on images from all institutions. The standard deviation of 10-fold cross-validation was 73 µm. The standard deviation for other institutions was reduced by re-learning. We describe the first application and validation of a deep learning approach for the estimation of central choroidal thickness from fundus images. This algorithm is expected to help graders judge choroidal thickening and thinning.

Atypical Focal Choroidal Excavation with Macular Hole in a Patient with Alagille Syndrome.

This study is aimed at reporting a rare and unusual focal choroidal excavation with a macular hole in a patient with Alagille syndrome (AGS). A 21-year-old woman with an established early-life AGS diagnosis was referred to our hospital prior to liver transplantation. Examination revealed best-corrected visual acuity of 16/20 and 20/20 in the right and left eye, respectively. Slit-lamp examination was positive for posterior embryotoxon in both eyes. Fundoscopy revealed diffuse choroidal hypopigmentation with increased visibility of the choroidal blood vessels and circumferential chorioretinal atrophy in the mid-peripheral and peripheral retina in both eyes. A full-thickness macular hole with underlying focal choroidal excavation was observed in the right eye. Optical coherence tomography through the macula confirmed choroidal excavation with a full-thickness macular hole in the right eye. To our knowledge, this is the first report describing focal choroidal excavation with a macular hole in an AGS patient.

Deep-learning-based AI for evaluating estimated nonperfusion areas requiring further examination in ultra-widefield fundus images.

We herein propose a PraNet-based deep-learning model for estimating the size of non-perfusion area (NPA) in pseudo-color fundus photos from an ultra-wide-field (UWF) image. We trained the model with focal loss and weighted binary cross-entropy loss to deal with the class-imbalanced dataset, and optimized hyperparameters in order to minimize validation loss. As expected, the resultant PraNet-based deep-learning model outperformed previously published methods. For verification, we used UWF fundus images with NPA and used Bland-Altman plots to compare estimated NPA with the ground truth in FA, which demonstrated that bias between the eNPA and ground truth was smaller than 10% of the confidence limits zone and that the number of outliers was less than 10% of observed paired images. The accuracy of the model was also tested on an external dataset from another institution, which confirmed the generalization of the model. For validation, we employed a contingency table for ROC analysis to judge the sensitivity and specificity of the estimated-NPA (eNPA). The results demonstrated that the sensitivity and specificity ranged from 83.3-87.0% and 79.3-85.7%, respectively. In conclusion, we developed an AI model capable of estimating NPA size from only an UWF image without angiography using PraNet-based deep learning. This is a potentially useful tool in monitoring eyes with ischemic retinal diseases.

Predicting the likelihood of need for future keratoplasty intervention using artificial intelligence.

OBJECTIVE: To apply artificial intelligence (AI) for automated identification of corneal condition and prediction of the likelihood of need for future keratoplasty intervention from optical coherence tomography (OCT)-based corneal parameters. DESIGN: Cohort study. PARTICIPANTS: We collected 12,242 corneal OCT images from 3162 subjects using CASIA OCT Imaging Systems (Tomey, Japan). We included 3318 measurements collected at the baseline visit of each patient. A total of 333 eyes had post-operative penetrating keratoplasty (PKP), lamellar keratoplasty (LKP), deep anterior keratoplasty (DALK), descemet's stripping automated endothelial keratoplasty (DSAEK) or descemet's membrane endothelial keratoplasty (DMEK) intervention. METHOD: We developed a pipeline including linear and nonlinear data transformations followed by unsupervised machine learning and applied on corneal parameters from the baseline visit of each patient. Five non-overlapping clusters of eyes were identified. Post hoc analyses revealed that clusters corresponded to different likelihoods of need for future keratoplasty. These clusters on a 2-dimensional map can be used by clinicians and surgeons to identify patients with higher risk of need for future keratoplasty intervention. MAIN OUTCOME MEASURES: The likelihood of the need for future surgery. RESULTS: The mean age of participants was 69.7 (standard deviation; SD = 16.1) and 59% were female. The normalized likelihood of need for future corneal keratoplasty intervention for eyes mapped onto clusters one to five were 2.2%, 1.0%, 33.1%, 32.7%, and 31.0%, respectively. CONCLUSIONS: The AI system can assist the (cornea) surgeon in identifying those patients who may be at higher risk for future keratoplasty using comprehensive corneal shape, thickness, and elevation parameters. Future research utilizing independent datasets is necessary to validate the proposed system.

Half-dose photodynamic therapy for serous non-neovascular retinal pigment epithelial detachment.

Purpose: To investigate the efficacy of half-dose photodynamic therapy (PDT) for treating symptomatic serous non-neovascular pigment epithelial detachment (PED) and to identify factors associated with treatment outcome. Patients and methods: This is a single-institute retrospective case series involving 12 eyes of 12 consecutive patients with serous non-neovascular PED who received half-dose PDT. We investigated the temporal change in best-corrected visual acuity (BCVA), the anatomic outcomes of PED, including recurrence, central choroidal thickness (CCT), and also choroidal thickness as a possible factor associated with treatment outcome. Results: At baseline, mean logMAR BCVA was 0.06 (range Snellen equivalent 6/20 to 24/20); 24 months later, it was 0.01 (Snellen equivalent 6/20 to 24/20; P=0.3). At 3 months, PED had completely flattened in 8, decreased in height in 2, and remained unchanged in 2. The CCT in eyes with PED was 410 µm and thicker than that in fellow eyes (290 µm, P<0.0001). Recurrence was noted in 2 within 12 months after treatment. Eyes in which the choroidal thickness with a baseline CCT >350µm or a 100 µm difference in CCT responded favorably to half-dose PDT (P=0.02). Conclusion: Half-dose PDT is effective in patients with non-neovascular PED. Thicker choroid and some difference in CCT between affected and fellow eye might be a good index for predicting the efficacy of this treatment.

Keratoconus severity identification using unsupervised machine learning.

We developed an unsupervised machine learning algorithm and applied it to big corneal parameters to identify and monitor keratoconus stages. A big dataset of corneal swept source optical coherence tomography (OCT) images of 12,242 eyes acquired from SS-1000 CASIA OCT Imaging Systems in multiple centers across Japan was assembled. A total of 3,156 eyes with valid Ectasia Status Index (ESI) between zero and 100% were selected for the downstream analysis. Four hundred and twenty corneal topography, elevation, and pachymetry parameters (excluding ESI Keratoconus indices) were selected. The algorithm included three major steps. 1) Principal component analysis (PCA) was used to linearly reduce the dimensionality of the input data from 420 to eight significant principal components. 2) Manifold learning was used to further reducing the selected principal components nonlinearly to two eigen-parameters. 3) Finally, a density-based clustering was applied to the eigen-parameters to identify eyes with keratoconus. Visualization of clusters in 2-D space was used to validate the quality of learning subjectively and ESI was used to assess the accuracy of the identified clusters objectively. The proposed method identified four clusters; I: a cluster composed of mostly normal eyes (224 eyes with ESI equal to zero, 23 eyes with ESI between five and 29, and nine eyes with ESI greater than 29), II: a cluster composed of mostly healthy eyes and eyes with forme fruste keratoconus (1772 eyes with ESI equal to zero, 698 eyes with ESI between five and 29, and 117 eyes with ESI greater than 29), III: a cluster composed of mostly eyes with mild keratoconus stage (184 eyes with ESI greater than 29, 74 eyes with ESI between five and 29, and 6 eyes with ESI equal to zero), and IV: a cluster composed of eyes with mostly advanced keratoconus stage (80 eyes had ESI greater than 29 and 1 eye had ESI between five and 29). We found that keratoconus status and severity can be well identified using unsupervised machine learning algorithms along with linear and non-linear corneal data transformation. The proposed method can better identify and visualize the keratoconus stages.

Applying artificial intelligence to disease staging: Deep learning for improved staging of diabetic retinopathy.

PURPOSE: Disease staging involves the assessment of disease severity or progression and is used for treatment selection. In diabetic retinopathy, disease staging using a wide area is more desirable than that using a limited area. We investigated if deep learning artificial intelligence (AI) could be used to grade diabetic retinopathy and determine treatment and prognosis. METHODS: The retrospective study analyzed 9,939 posterior pole photographs of 2,740 patients with diabetes. Nonmydriatic 45° field color fundus photographs were taken of four fields in each eye annually at Jichi Medical University between May 2011 and June 2015. A modified fully randomly initialized GoogLeNet deep learning neural network was trained on 95% of the photographs using manual modified Davis grading of three additional adjacent photographs. We graded 4,709 of the 9,939 posterior pole fundus photographs using real prognoses. In addition, 95% of the photographs were learned by the modified GoogLeNet. Main outcome measures were prevalence and bias-adjusted Fleiss' kappa (PABAK) of AI staging of the remaining 5% of the photographs. RESULTS: The PABAK to modified Davis grading was 0.64 (accuracy, 81%; correct answer in 402 of 496 photographs). The PABAK to real prognosis grading was 0.37 (accuracy, 96%). CONCLUSIONS: We propose a novel AI disease-staging system for grading diabetic retinopathy that involves a retinal area not typically visualized on fundoscopy and another AI that directly suggests treatments and determines prognoses.

An unusual presentation of concurrent congenital grouped pigmentation of the retina with albinotic spots.

A 10-year-old Japanese girl presented with multiple, small, flat, dark brown clusters of retinal pigment epithelium in both eyes. She had no other symptoms. The appearance was consistent with congenital grouped pigmentation of the retina or bear track spots. A unique co-existing feature was the presence of small, albinotic spots within the entire retina. Fluorescein angiography demonstrated a persistent hypofluorescence correlating with the grouped pigmentation of the retina and granular hyperfluorescence correlating with the albinotic spots. Her mother also showed irregular pigmentation and albinotic spots in both eyes. We describe an extremely rare case of congenital grouped pigmentation and albinotic spots occurring in the retina of the same individual.

Acute macular neuroretinopathy in a 15-year-old boy: optical coherence tomography and visual acuity findings.

PURPOSE: To report a case of acute macular neuroretinopathy (AMN) in a 15-year-old boy. METHODS: Images were obtained with fundus photography and optical coherence tomography (OCT). RESULTS: The patient complained of blurred vision and a small central scotoma in the left eye. Left visual acuity was 0.3. Fundus photographs revealed a small dark area in the fovea of the left eye. OCT showed attenuation of the photoreceptor inner segment (IS)/outer segment (OS) line and the OS/retinal pigment epithelium (RPE) line in the left eye. One week after the initial visit, left visual acuity had improved to 0.6, and there was partial recovery of the IS/OS line with focal thinning of the OS/RPE line. One month later, left visual acuity had improved to 0.8, and OCT showed complete restoration of the IS/OS line and the OS/RPE line. CONCLUSIONS: To our knowledge, this is the youngest patient reported with a diagnosis of AMN. Changes in microstructural findings and visual acuity were analogous during recovery of AMN in our patient.

Ocular findings in two siblings with Joubert syndrome.

We describe two brothers with Joubert syndrome (JS). JS diagnosis was made on the basis of neurological findings and the presence of the characteristic "molar tooth sign", which was subsequently confirmed by magnetic resonance imaging. Both brothers demonstrated ptosis, hypotropia, exotropia, and horizontal pendular nystagmus. The younger brother had mild chorioretinal discoloration at the peripapillary region in both eyes, and a small coloboma at the inferior region of his right optic disc. The elder brother had coloboma in his right eye and a colobomatous optic disc in his left eye. Optical coherence tomography showed that the foveal architecture was preserved in both patients. We discuss the ocular findings, including those from optical coherence tomography, in JS, which has recently been recognized as ciliopathy.

Correlations between choroidal abnormalities, Lisch nodules, and age in patients with neurofibromatosis type 1.

To evaluate correlations between choroidal abnormalities, Lisch nodules, and age in patients with neurofibromatosis type 1 (NF1), we examined ten cases with NF1 using near-infrared reflectance imaging. Patients ranged in age from 4 to 39 years. The angle used for near-infrared reflectance imaging was 55°. We counted the total number of choroidal abnormalities in an area within a 55° angle centered on the fovea and the total number of Lisch nodules on the iris by slit-lamp examination. No positive correlation was found between the number of Lisch nodules and patient age (Spearman's rank correlation coefficient ρ=0.117, P=0.7414). Choroidal abnormalities tended to increase with age (ρ=0.6150), but this difference was not statistically significant (P=0.0650). A positive correlation was found between the number of choroidal abnormalities and Lisch nodules (ρ=0.783, P=0.0267). In conclusion, choroidal abnormalities tend to increase with patient age and are correlated with the number of Lisch nodules.

Rare and unusual choroidal abnormalities in a patient with systemic lupus erythematosus.

PURPOSE: To report a case of rare and unusual choroidal abnormalities in a 42-year-old woman with systemic lupus erythematosus (SLE). METHODS: Images were obtained using fundus photography, fluorescein angiography, near-infrared reflectance (NIR) imaging, and optical coherence tomography (OCT). RESULTS: The patient had a history of SLE and central retinal artery occlusion in her right eye. Fundus examination showed no specific retinochoroidal abnormalities, with the exception of optic disc atrophy in her right eye and a peripapillary small hemorrhage in her left eye. However, NIR revealed multiple bright patchy lesions in the choroid of the posterior pole and the mid-periphery of the fundus in both eyes. OCT demonstrated irregular hyperreflectivity at the lesion sites. CONCLUSIONS: The observed choroidal abnormalities are highly specific findings and therefore indicative of neurofibromatosis type 1 (NF1). Since the coexistence of SLE and NF1 is extremely rare, this case provided the chance to examine the relationship between SLE and NF1.

Optical coherence tomography imaging of choroidal abnormalities in neurofibromatosis type 1.

We report a case of neurofibromatosis type 1 (NF1) examined by infrared fundus autofluorescence (IR-FAF) and optical coherence tomography (OCT) to characterize the associated choroidal abnormalities. The conventional ophthalmoscopic findings were unremarkable. However, IR-FAF revealed multiple bright patchy lesions in the choroid of the posterior pole, in both eyes. OCT demonstrated irregular hyperreflectivity at the sites of these lesions. Patients with NF1 may have typical choroidal lesions that are visible on IR-FAF, which can be confirmed through OCT.

Changes in optical coherence tomography findings in acute zonal occult outer retinopathy.

PURPOSE: To report a case of acute zonal occult outer retinopathy (AZOOR) in a 39-year-old woman. METHODS: Images were obtained with fundus photography and optical coherence tomography (OCT). A multifocal electroretinogram (mfERG) was also obtained to evaluate retinal function. RESULTS: The patient's right visual acuity was 0.8. Fundus photographs showed no specific abnormal findings. OCT showed attenuation of the photoreceptor inner segment/outer segment (IS/OS) line and the cone outer segment tip (COST) line in the right eye. The mfERG showed corresponding amplitude reductions. One month after the initial visit, her right visual acuity improved to 1.2. There was partial recovery of the IS/OS line; however, the COST line was still absent, and there was no apparent improvement in the mfERG responses. CONCLUSIONS: Our findings showed a discrepancy between the microstructural findings and visual function during recovery of AZOOR.

An optical coherence tomography and fundus autofluorescence imaging study of peripapillary acute zonal occult outer retinopathy.

PURPOSE: To evaluate peripapillary acute zonal occult outer retinopathy in a 67-year-old man. METHODS: Images were obtained using fundus photography, fluorescein angiography (FA) and indocyanine green angiography (ICGA), fundus autofluorescence (FAF) imaging, and optical coherence tomography (OCT). Visual field testing and multifocal electroretinogram (mfERG) were also performed to evaluate retinal function. RESULTS: Fundus photographs showed subtle pigmentary changes in the peripapillary region, while FAF imaging showed clearly defined hypofluorescent areas in the peripapillary region. Intense hyperfluorescent lesions were also seen underneath hypofluorescent areas. A transmission defect with a granular hyperfluorescence was visible on FA, and ICGA showed hypofluorescence within the lesion. The outer border of the peripapillary zone appeared to block the underlying choroidal fluorescence. The photoreceptor inner segment/outer segment line was absent on OCT images from this area in both eyes, and hyperreflective punctate drusen-like materials were present at the outer border of the peripapillary zone. Visual field testing by Goldmann perimetry showed blind-spot enlargement, and mfERG showed corresponding amplitude reductions. CONCLUSIONS: We speculate that the intense hyperautofluorescent material at the outer border of the peripapillary zone might be accumulated drusenoid lipofuscin.

Prepapillary vascular loop associated with persistent hyperplastic primary vitreous.

A 66-year-old woman was referred for evaluation of retinal vessel abnormality. She had no visual symptoms. The anterior segment showed a retrolental opacity in her right eye. Fundus examination showed a bilateral prepapillary vascular loop associated with PHPV in her right eye. Prepapillary vascular loops were present in both eyes, although they were much more prominent in the right eye. To our knowledge, there are no reports of prepapillary vascular loops associated with PHPV.