A Preliminary Evaluation of the Diagnostic Performance of a Smartphone-Based Machine Learning-Assisted System for Evaluation of Clinical Activity Score in Digital Images of Thyroid-Associated Orbitopathy.

Background: We previously developed a machine learning (ML)-assisted system for predicting the clinical activity score (CAS) in thyroid-associated orbitopathy (TAO) using digital facial images taken by a digital single-lens reflex camera in a studio setting. In this study, we aimed to apply this system to smartphones and detect active TAO (CAS ≥3) using facial images captured by smartphone cameras. We evaluated the performance of our system on various smartphone models and compared it with the performance of ophthalmologists with varying clinical experience. Methods: We applied the preexisting ML architecture to classify photos taken with smartphones (Galaxy S21 Ultra, iPhone 12 pro, iPhone 11, iPhone SE 2020, Galaxy M20, and Galaxy A21S). The performance was evaluated with smartphone-captured images from 100 patients with TAO. Three ophthalmology residents, three general ophthalmologists with <5 years of clinical experience, and three oculoplastic specialists independently interpreted the same set of images taken under a studio environment and compared their results with those generated by the smartphone-based ML-assisted system. Reference CAS was determined by a consensus of three oculoplastic specialists. Results: Active TAO (CAS ≥3) was identified in 28 patients. Smartphone model used in capturing facial images influenced active TAO detection performance (F1 score 0.59-0.72). The smartphone-based system showed 74.5% sensitivity, 84.8% specificity, and F1 score 0.70 on top three smartphones. On images from all six smartphones, average sensitivity, specificity, and F1 score were 71.4%, 81.6%, and 0.66, respectively. Ophthalmology residents' values were 69.1%, 55.1%, and 0.46. General ophthalmologists' values were 61.9%, 79.6%, and 0.55. Oculoplastic specialists' values were 73.8%, 90.7%, and 0.75. This smartphone-based ML-assisted system predicted CAS within 1 point of reference CAS in 90.7% using facial images from smartphones. Conclusions: Our smartphone-based ML-assisted system shows reasonable accuracy in detecting active TAO, comparable with oculoplastic specialists and outperforming residents and general ophthalmologists. It may enable reliable self-monitoring for disease activity, but confirmatory research is needed for clinical application.

Quantitative microvascular analysis in different stages of retinitis pigmentosa using optical coherence tomography angiography.

As retinitis pigmentosa (RP) is chronic and progressive, the chronological sequence of microvascular changes is important for understanding its pathophysiology. We aimed to investigate retinal and choroidal microvascular changes according to the RP stages. The stages of RP were classified into three stages according to the integrity and width of the inner segment ellipsoid zone: early, ≥ 2500 µm; moderate, < 2500 µm; advanced, absence. Using optical coherence tomography angiography, quantitative microvascular parameters were analyzed. In total, 91 eyes from 49 patients were included. For the superficial capillary plexus (SCP) and deep capillary plexus (DCP), perfusion densities (PDs) in the early stage (SCP: 37.32 ± 8.11%; DCP: 21.19 ± 9.15%) were greater than those in moderate (SCP: 34.16 ± 6.65%, P = 0.011; DCP: 15.67 ± 8.85%, P = 0.031) and advanced stages (SCP: 33.71 ± 9.02%, P = 0.030; DCP: 12.83 ± 6.29%, P < 0.001). The choroidal vascularity index in the early stage (0.58 ± 0.03) was greater than those in the moderate (0.57 ± 0.02, P = 0.017) and advanced stage (0.56 ± 0.02, P = 0.033). The area and perimeter of foveal avascular zone (FAZ) in advanced stage (0.44 ± 0.26 mm2, 2.96 ± 0.86 mm, respectively) were larger than those in early (0.26 ± 0.11 mm2, P = 0.020; 2.19 ± 0.53 mm, P = 0.006, respectively) and moderate stage (0.28 ± 0.13 mm2, P = 0.043; 2.24 ± 0.67 mm, P = 0.013, respectively). During RP disease progression, retinal and choroidal microvascular vessel density decreases in the early stage, followed by FAZ enlargement in the advanced stage.

Longer axial length is associated with better prediction for refractive error after sutureless flanged intrascleral fixation of intraocular lens.

OBJECTIVE: To investigate refractive outcomes and associated factors after sutureless flanged intrascleral fixation of intraocular lens (SFIF-IOL). METHODS: We retrospectively reviewed the medical records of consecutive patients who underwent SFIF-IOL at a single centre. The prediction error (PE; difference between the achieved and target refractive error) and absolute PE (APE) were analysed. Risk factors associated with refractive surprise, defined as APE > +0.5 D, were investigated using multivariable logistic regression analysis. RESULTS: Ninety-one eyes were included. At the final follow-up, the mean PE and APE were +0.07 ± 0.88 and +0.68 ± 0.56 D, respectively. Refractive surprise was observed in 44 eyes (54.3%) and was associated with a shorter axial length (AL) [odds ratio, 0.825; 95% confidence interval, 0.688-0.991; P = 0.039]. APE showed a significant correlation with AL at the final visit (⍴ = -0.269, P = 0.010), and eyes with AL ≥ 26 mm had significantly lower APE than did those with AL of 24-26 mm (P = 0.021) and AL < 24 mm (P = 0.0059). CONCLUSIONS: The refractive outcome after SFIF-IOL using manufacturer's A constant was favourable on average. Eyes with a longer AL were more likely to show a smaller deviation from the target refraction.

Whole genome sequencing for inherited retinal diseases in the Korean National Project of Bio Big Data.

PURPOSE: This study aimed to analyze the genetic results of inherited retinal diseases (IRDs) and evaluate the diagnostic usefulness of whole genome sequencing (WGS) in the Korean National Project of Bio Big Data. METHODS: As part of the Korean National Project of Bio Big Data, WGS was performed on 32 individuals with IRDs with no identified pathogenic variants through whole or targeted exome sequencing. RESULTS: Individuals with retinitis pigmentosa (n = 23), cone dystrophy (n = 2), cone-rod dystrophy (n = 2), familial exudative vitreoretinopathy (n = 2), pigmented paravenous chorioretinal atrophy (n = 1), North Carolina macular dystrophy (n = 1), and bull's-eye macular dystrophy (n = 1) were included. WGS revealed genetic mutations in the IQCB1, PRPF31, USH2A, and GUCY2D genes in five cases (15.6%). Two large structural variations and an intronic variant were newly detected in three cases. Two individuals had biallelic missense mutations that were not identified in previous exome sequencing. CONCLUSION: With WGS, the causative variants in 15.6% of unsolved IRDs from the Korean National Project of Bio Big Data were identified. Further research with a larger cohort might unveil the diagnostic usefulness of WGS in IRDs and other diseases.

Accuracy of the PEARL-DGS Formula for Intraocular Lens Power Calculation in Post-Myopic Laser Refractive Corneal Surgery Eyes.

PURPOSE: To investigate the accuracy of the PEARL-DGS formula for intraocular lens (IOL) power calculation in post-myopic laser refractive corneal surgery eyes. DESIGN: Retrospective case series. METHODS: A total of 139 eyes of 139 patients (mean axial length: 27.4 ± 2.1 mm) who had prior myopic laser refractive corneal surgery and subsequent cataract surgery using Tecnis ZCB00 from March 2018 to February 2023 were included. Refractive outcomes of 5 formulas (Barrett True K, Haigis-L, Hoffer-QST, PEARL-DGS, and Shammas-PL) were evaluated. Prediction error was defined as the difference between the measured and predicted postoperative refractive spherical equivalent using the IOL power actually implanted. Mean prediction error (MPE), median absolute prediction error (MedAE), and mean absolute prediction error were calculated. RESULTS: Without constant optimization, the PEARL-DGS resulted in a MPE of +0.05 ± 0.65 diopters (D), whereas the other formulas resulted in myopic shifts. The MedAEs of the formulas were 0.39, 0.53, 0.65, 0.85, and 1.11 D for the PEARL-DGS, Hoffer-QST, Barrett True K, Shammas-PL, and Haigis-L, respectively, in order of magnitude (P < .05). With constant optimization, there were no statistically significant differences in the MedAEs among the 5 formulas (P = .388). CONCLUSIONS: In comparison to other IOL formulas, the PEARL-DGS resulted in better refractive outcomes after cataract surgery in post-myopic laser refractive corneal surgery eyes without constant optimization. We suggest that PEARL-DGS be considered as the first choice for IOL power calculation in these eyes when the clinicians do not have their optimized constants.

Deep Learning-based Prediction of Axial Length Using Ultra-widefield Fundus Photography.

PURPOSE: To develop a deep learning model that can predict the axial lengths of eyes using ultra-widefield (UWF) fundus photography. METHODS: We retrospectively enrolled patients who visited the ophthalmology clinic at the Seoul National University Hospital between September 2018 and December 2021. Patients with axial length measurements and UWF images taken within 3 months of axial length measurement were included in the study. The dataset was divided into a development set and a test set at an 8:2 ratio while maintaining an equal distribution of axial lengths (stratified splitting with binning). We used transfer learning-based on EfficientNet B3 to develop the model. We evaluated the model's performance using mean absolute error (MAE), R-squared (R2), and 95% confidence intervals (CIs). We used vanilla gradient saliency maps to illustrate the regions predominantly used by convolutional neural network. RESULTS: In total, 8,657 UWF retinal fundus images from 3,829 patients (mean age, 63.98 ±15.25 years) were included in the study. The deep learning model predicted the axial lengths of the test dataset with MAE and R2 values of 0.744 mm (95% CI, 0.709-0.779 mm) and 0.815 (95% CI, 0.785-0.840), respectively. The model's accuracy was 73.7%, 95.9%, and 99.2% in prediction, with error margins of ±1.0, ±2.0, and ±3.0 mm, respectively. CONCLUSIONS: We developed a deep learning-based model for predicting the axial length from UWF images with good performance.

SS-OCT-based ocular biometry in an adult Korean population with cataract.

PURPOSE: To evaluate the characteristics of ocular biometric parameters in adult Korean patients with cataract. SETTING: Department of Ophthalmology, Seoul National University Hospital, Seoul, South Korea. DESIGN: Retrospective case series. METHODS: Ocular biometric values of 5273 eyes of 5273 Korean patients undergoing cataract surgery measured with the IOLMaster 700 at the Seoul National University Hospital between November 2019 and December 2021 were reviewed. RESULTS: A total of 5273 eyes of 5273 Korean patients were analyzed. The mean ± SD age was 66.1 ± 12.8 years, and 62% were female. Overall, age and ocular biometric parameters were correlated with each other. Particularly, age showed a negative correlation with anterior chamber depth (ACD; r = -0.357), axial length (AL; r = -0.344), and posterior keratometry (PK; r = -0.054) and a positive correlation with lens thickness (LT; r = 0.484), angle α ( r = 0.194), total keratometry (TK; r = 0.137), anterior keratometry (AK; r = 0.129), and angle κ ( r = 0.071). AL showed a positive correlation with ACD ( r = 0.503) and PK ( r = 0.339) and a negative correlation with AK ( r = -0.342), TK ( r = -0.334), LT ( r = -0.288), angle α ( r = -0.220), and angle κ ( r = -0.040). With age, anterior and total corneal astigmatism changed from with-the-rule (WTR) to against-the-rule (ATR) astigmatism. Posterior corneal astigmatism was ATR regardless of age; however, the magnitude decreased with age. CONCLUSIONS: Age showed a significant correlation in the order of LT, ACD, and AL. With age, angle α and κ increased, and total corneal astigmatism changes from WTR to ATR, which is mainly affected by changes in anterior corneal astigmatism. AL showed a significant correlation in the order of ACD, AK, PK, and TK. These data are pertinent for improving the result after cataract surgery, especially when using premium intraocular lenses.

Hemisphere opposite to central retinal vascular trunk deviation is earlier affected by glaucomatous damage in primary angle-closure glaucoma.

PURPOSE: The purpose of the study was to investigate whether the position of the central retinal vascular trunk (CRVT), as a surrogate for lamina cribrosa (LC) offset, is associated with the dominant hemisphere of visual field defect in primary angle-closure glaucoma (PACG) eyes. METHODS: Central retinal vascular trunk deviation was measured from Bruch's membrane opening (BMO) centre, which was delineated by OCT imaging, using the horizontal midline as a reference. The dominant hemisphere developing visual field defect was defined as three connected abnormal points (having a p-value < 5% probability of being normal) appearing in only one hemisphere or each point of the hemisphere having a statistically worse value compared with its mirrored point in the opposite hemisphere on pattern deviation plots. RESULTS: One hundred five (80%) of 132 eyes with PACG had dominant hemisphere of visual field defect initially: 70 eyes (67%) in the superior and 35 eyes (33%) in the inferior hemisphere. The CRVT was located superiorly in the dominant superior visual field defect group (p < 0.001). A logistic regression analysis revealed that superior deviation of the CRVT was the only factor associated with dominant superior visual field defect (p < 0.001). Externally oblique border (EOB) presence was associated with larger BMO (p = 0.005) and angular deviation of CRVT (p = 0.002). CONCLUSIONS: Central retinal vascular trunk deviation was associated with the dominant hemisphere of visual field defect in PACG eyes. This finding implies that the LC position relative to the BMO centre (LC/BMO offset) may incur structural vulnerability in the optic nerve head of PACG eyes.

DETECTION AND LOCALIZATION OF RETINAL BREAKS IN ULTRAWIDEFIELD FUNDUS PHOTOGRAPHY USING a YOLO v3 ARCHITECTURE-BASED DEEP LEARNING MODEL.

PURPOSE: We aimed to develop a deep learning model for detecting and localizing retinal breaks in ultrawidefield fundus (UWF) images. METHODS: We retrospectively enrolled treatment-naive patients diagnosed with retinal break or rhegmatogenous retinal detachment and who had UWF images. The YOLO v3 architecture backbone was used to develop the model, using transfer learning. The performance of the model was evaluated using per-image classification and per-object detection. RESULTS: Overall, 4,505 UWF images from 940 patients were used in the current study. Among them, 306 UWF images from 84 patients were included in the test set. In per-object detection, the average precision for the object detection model considering every retinal break was 0.840. With the best threshold, the overall precision, recall, and F1 score were 0.6800, 0.9189, and 0.7816, respectively. In the per-image classification, the model showed an area under the receiver operating characteristic curve of 0.957 within the test set. The overall accuracy, sensitivity, and specificity in the test data set were 0.9085, 0.8966, and 0.9158, respectively. CONCLUSION: The UWF image-based deep learning model evaluated in the current study performed well in diagnosing and locating retinal breaks.

Comparison of ocular biometric measurements in patients with cataract using three swept-source optical coherence tomography devices.

BACKGROUND: Precise measurement of ocular biometry is critical for determining intraocular lens power. Newly developed swept-source optical coherence tomography (SS-OCT) - based ocular biometric devices, ANTERION and CASIA2 provide ocular biometric measurements as IOLMaster 700. This study aimed to assess agreement between three devices. METHODS: This retrospective comparative study includes patients with cataract who underwent ocular biometric measurements with three devices, ANTERION, CASIA2, and IOLMaster 700, at Seoul National University Hospital, in April 2020. Anterior keratometry, total keratometry, central corneal thickness (CCT), anterior chamber depth (ACD), lens thickness (LT), and axial length (AL) were the main parameters for the comparison. To assess the agreement between the devices, intraclass coefficient (ICC) and Bland-Altman analysis with 95% limits of agreement (LoA) were used. RESULTS: A total of 47 eyes of 29 patients were measured with three devices. Average anterior keratometry showed excellent agreement (ICC ≥ 0.989), and the mean difference was less than 0.1 D. However, the ICC of the total average keratometry ranged from 0.808 to 0.952, and the difference was more than 0.43 D. The AL measured by ANTERION and IOLMaster 700 showed excellent agreement (ICC = 0.999), and the mean difference was 0.005 mm. The ANTERION and IOLMaster 700 did not obtain AL in six (12.8%) and three (6.4%) cases, respectively (P = 0.001 by Fisher's exact test). The CCT, ACD, and LT also showed excellent agreement (ICC > 0.9). CONCLUSIONS: The new SS-OCT-based devices, ANTERION, and CASIA2 showed a good agreement with IOLMaster 700 in measuring ocular biometry except for the total keratometry. The AL of ANTERION and IOLMaster 700 showed excellent agreement.

Optical coherence tomography angiographic findings of lamellar macular hole: comparisons between tractional and degenerative subtypes.

We investigated the microvascular changes in eyes with lamellar macular holes (LMHs) using optical coherence tomography angiography (OCTA), compare them between two subtypes of LMH. Tractional and degenerative LMH were differentiated based on the morphological characteristics of OCT. In OCTA images, foveal and parafoveal vessel density (VD) in the superficial and deep capillary plexus (SCP, DCP) and foveal avascular zone (FAZ) area were measured. Eyes that underwent vitrectomy for LMH were included in subgroup analysis. We analysed 63 LMH (42 tractional and 21 degenerative) eyes and 63 control eyes. Compared with degenerative LMH, tractional LMH had better BCVA (p = 0.010), smaller FAZ area (p = 0.001), and higher foveal VD in the SCP (p = 0.130) and DCP (p = 0.002). In degenerative LMH, better BCVA was associated with greater foveal VD in the SCP (p = 0.040) and DCP (p = 0.005), and parafoveal VD in the SCP (p = 0.006). In subgroup analysis, only the tractional LMH group showed significant increases in foveal and parafoveal VDs in the SCP after vitrectomy (p = 0.001 and p = 0.026, respectively). Significant differences in microvascular changes were noted between tractional and degenerative LMH, suggesting that two subtypes are distinct pathogenetic entities.

Choroidal Neovascularization Secondary to Central Serous Chorioretinopathy: OCT Angiography Findings and Risk Factors.

PURPOSE: To identify the clinical characteristics and risk factors for secondary choroidal neovascularization (CNV) in central serous chorioretinopathy (CSC). METHODS: In this retrospective study, we included a total of 108 eyes in 106 CSC patients. Group A was defined as patients initially diagnosed with CSC who developed secondary CNV, and group B was defined as patients who did not develop secondary CNV. Clinical and demographic characteristics, optical coherence tomography (OCT) findings at CSC diagnosis and OCT angiography (OCTA) at the time of secondary CNV diagnosis, were compared between the groups. RESULTS: Thirty-one eyes had CNV (group A) and 77 eyes did not (group B). The mean age of group A was higher than that of group B (52.28 ± 6.87 vs. 46.78 ± 9.45 years; P < 0.001). Although there was no difference in pigment epithelial detachment (PED) height, group A had larger PED width than group B at CSC diagnosis. The foveal and parafoveal choriocapillary flow densities were significantly lower in group A than group B (P < 0.001). Although there was no difference in pigment epithelial detachment (PED) height, group A had larger PED width than group B at CSC diagnosis. The foveal and parafoveal choriocapillary flow densities were significantly lower in group A than group B (P < 0.001). Although there was no difference in pigment epithelial detachment (PED) height, group A had larger PED width than group B at CSC diagnosis. The foveal and parafoveal choriocapillary flow densities were significantly lower in group A than group B (. CONCLUSION: We identified that older age, wider PED width at diagnosis, and recurrent episodes of CSC were independent risk factors for development of secondary CNV. Therefore, patients with these risk factors should be monitored to allow early detection and prompt treatment of secondary CNV.