A fundus image dataset for intelligent retinopathy of prematurity system.

Image-based artificial intelligence (AI) systems stand as the major modality for evaluating ophthalmic conditions. However, most of the currently available AI systems are designed for experimental research using single-central datasets. Most of them fell short of application in real-world clinical settings. In this study, we collected a dataset of 1,099 fundus images in both normal and pathologic eyes from 483 premature infants for intelligent retinopathy of prematurity (ROP) system development and validation. Dataset diversity was visualized with a spatial scatter plot. Image classification was conducted by three annotators. To the best of our knowledge, this is one of the largest fundus datasets on ROP, and we believe it is conducive to the real-world application of AI systems.

Automated detection of nine infantile fundus diseases and conditions in retinal images using a deep learning system.

Purpose: We developed an Infant Retinal Intelligent Diagnosis System (IRIDS), an automated system to aid early diagnosis and monitoring of infantile fundus diseases and health conditions to satisfy urgent needs of ophthalmologists. Methods: We developed IRIDS by combining convolutional neural networks and transformer structures, using a dataset of 7697 retinal images (1089 infants) from four hospitals. It identifies nine fundus diseases and conditions, namely, retinopathy of prematurity (ROP) (mild ROP, moderate ROP, and severe ROP), retinoblastoma (RB), retinitis pigmentosa (RP), Coats disease, coloboma of the choroid, congenital retinal fold (CRF), and normal. IRIDS also includes depth attention modules, ResNet-18 (Res-18), and Multi-Axis Vision Transformer (MaxViT). Performance was compared to that of ophthalmologists using 450 retinal images. The IRIDS employed a five-fold cross-validation approach to generate the classification results. Results: Several baseline models achieved the following metrics: accuracy, precision, recall, F1-score (F1), kappa, and area under the receiver operating characteristic curve (AUC) with best values of 94.62% (95% CI, 94.34%-94.90%), 94.07% (95% CI, 93.32%-94.82%), 90.56% (95% CI, 88.64%-92.48%), 92.34% (95% CI, 91.87%-92.81%), 91.15% (95% CI, 90.37%-91.93%), and 99.08% (95% CI, 99.07%-99.09%), respectively. In comparison, IRIDS showed promising results compared to ophthalmologists, demonstrating an average accuracy, precision, recall, F1, kappa, and AUC of 96.45% (95% CI, 96.37%-96.53%), 95.86% (95% CI, 94.56%-97.16%), 94.37% (95% CI, 93.95%-94.79%), 95.03% (95% CI, 94.45%-95.61%), 94.43% (95% CI, 93.96%-94.90%), and 99.51% (95% CI, 99.51%-99.51%), respectively, in multi-label classification on the test dataset, utilizing the Res-18 and MaxViT models. These results suggest that, particularly in terms of AUC, IRIDS achieved performance that warrants further investigation for the detection of retinal abnormalities. Conclusions: IRIDS identifies nine infantile fundus diseases and conditions accurately. It may aid non-ophthalmologist personnel in underserved areas in infantile fundus disease screening. Thus, preventing severe complications. The IRIDS serves as an example of artificial intelligence integration into ophthalmology to achieve better outcomes in predictive, preventive, and personalized medicine (PPPM / 3PM) in the treatment of infantile fundus diseases. Supplementary Information: The online version contains supplementary material available at 10.1007/s13167-024-00350-y.

An Artificial Intelligence System for Screening and Recommending the Treatment Modalities for Retinopathy of Prematurity.

PURPOSE: The purpose of this study was to develop an artificial intelligence (AI) system for the identification of disease status and recommending treatment modalities for retinopathy of prematurity (ROP). METHODS: This retrospective cohort study included a total of 24,495 RetCam images from 1075 eyes of 651 preterm infants who received RetCam examination at the Shenzhen Eye Hospital in Shenzhen, China, from January 2003 to August 2021. Three tasks included ROP identification, severe ROP identification, and treatment modalities identification (retinal laser photocoagulation or intravitreal injections). The AI system was developed to identify the 3 tasks, especially the treatment modalities of ROP. The performance between the AI system and ophthalmologists was compared using extra 200 RetCam images. RESULTS: The AI system exhibited favorable performance in the 3 tasks, including ROP identification [area under the receiver operating characteristic curve (AUC), 0.9531], severe ROP identification (AUC, 0.9132), and treatment modalities identification with laser photocoagulation or intravitreal injections (AUC, 0.9360). The AI system achieved an accuracy of 0.8627, a sensitivity of 0.7059, and a specificity of 0.9412 for identifying the treatment modalities of ROP. External validation results confirmed the good performance of the AI system with an accuracy of 92.0% in all 3 tasks, which was better than 4 experienced ophthalmologists who scored 56%, 65%, 71%, and 76%, respectively. CONCLUSIONS: The described AI system achieved promising outcomes in the automated identification of ROP severity and treatment modalities. Using such algorithmic approaches as accessory tools in the clinic may improve ROP screening in the future.

Twelve-year outcomes of bedside laser photocoagulation for severe retinopathy of prematurity.

Purpose: The purpose of this study is to evaluate the 12-year outcomes of bedside laser photocoagulation (LP) for severe retinopathy of prematurity (ROP) under sedation combined with ocular surface anesthesia in neonatal intensive care units (NICU). Design: The study is a retrospective case series. Methods: Infants treated with bedside LP for severe ROP from April 2009 to September 2021 were included. All LP treatments were performed under sedation and surface anesthesia at the bedside in NICU. Data were recorded for clinical and demographic characteristics, total laser spots, duration of treatment, proportion of total regression of ROP, proportion of recurrence, and adverse events. Results: A total of 364 infants (715 eyes) were included, with a mean gestational age of 28.6 ± 2.4 weeks (range: 22.6-36.6 weeks) and a mean birth weight of 1,156.0 ± 339.0 g (range: 480-2,200 g). The mean number of laser spots was 832 ± 469, and the mean duration of treatment was 23.5 ± 5.3 min per eye. Of all the eyes, 98.3% responded to LP with complete regression of ROP. ROP recurred in 15 (2.1%) eyes after the initial LP. Additional LP was performed in seven (1.0%) eyes. No patient exhibited mistaken LP of other ocular tissues, and there were no serious ocular adverse effects. None of them needed endotracheal intubation. Conclusions: Bedside LP treatment is effective and safe for premature infants with severe ROP under sedation and surface anesthesia in NICU, especially for infants whose general condition is unstable and not suitable for transport.

Awareness, Prevalence, and Knowledge of Dry Eye Among Internet Professionals: A Cross-sectional Study in China.

OBJECTIVES: To investigate awareness, prevalence, and knowledge of dry eye among Internet professionals in China. METHODS: A cross-sectional study was conducted among 1,265 randomly selected Internet professionals aged ≥18 years. A self-administered questionnaire was used to assess dry eye awareness, dry eye symptoms, and knowledge about dry eye risk factors. Data on demographics and complete medical history were also collected. The primary outcome was the rate of dry eye awareness determined by the answer to the question "Have you seen or heard anything about dry eye recently?" RESULTS: Of the 1,265 included individuals aged 20 to 49 years, 519 (41.0%) were women. 54.4% (688 of 1,265) of participants had seen or heard something about dry eye recently and most had obtained information through Internet. 50.8% (643 of 1,265) of participants were identified as subjects with symptoms of dry eye. Dry eye awareness was greater in contact lens wearers (odds ratio [OR], 6.49; 95% confidence interval [CI], 3.70-11.38; P <0.001), those with a refractive surgical history (OR, 5.09; 95% CI, 2.34-11.08; P <0.001), relatives and/or friends of ophthalmologists (OR, 2.76; 95% CI, 1.39-5.49; P =0.004), those with symptoms of dry eye (OR, 1.87; 95% CI, 1.47-2.38; P <0.001) and female subjects (OR, 1.44; 95% CI, 1.13-1.86; P =0.004). Knowledge of nonmodifiable and modifiable risk factors for dry eye was poor in substantial numbers of the participants. CONCLUSIONS: The level of dry eye awareness and knowledge of its risk factors is suboptimal in Internet professionals, although the Internet professionals are at high risk of the disease.

Study of the Biological Developmental Characteristics of the Eye in Children After Laser Surgery for the Treatment of Retinopathy of Prematurity.

OBJECTIVE: To observe the differences in ocular biology between premature infants who had undergone retinal laser photocoagulation (LP) for retinopathy of prematurity (ROP) and full-term infants and to investigate the relationships between these differences and the development of the refractive state. METHODS: This retrospective, cross-sectional study included 25 children (50 eyes) who had undergone laser treatment for aggressive posterior retinopathy of prematurity (AP-ROP), ROP in zone I requiring treatment, or ROP in zone II requiring treatment in the posterior pole (laser group) and 29 full-term infants (58 eyes) who had not (control group). Basic information, spherical equivalent (SE), and best corrected visual acuity (BCVA) were collected from the two groups. Their mean ages were 7.32 ± 2.85 and 7.34 ± 2.57 years, respectively (t = -0.047, P = 0.96). Ocular biology data were measured using an IOL Master 700 instrument (Carl Zeiss Meditec AG) and the data were processed using MATLAB (R2016a, Mathworks Inc.). The data markers included central corneal thickness (CCT), anterior and posterior surface corneal curvature radius (CCR), anterior chamber depth (ACD), lens thickness (LT), lens anterior surface curvature radius, lens posterior surface curvature radius, and eye axis length (AL). Optometric data were collected simultaneously and all BCVA values were converted to the logarithm of the minimum angle of resolution (LogMAR) for analysis. The data were statistically analyzed using SPSS software (V.23.0). Independent sample t-tests were used for the assessment of ocular biology and refractive indices in both groups of children and Pearson correlation coefficients were used to evaluate the correlations between age, gestational age at birth and ocular biology structural parameters. P < 0.05 was considered statistically significant. RESULTS: Comparisons of ocular biomarkers, refractive status, and BCVA between children in the laser and control groups showed relationships among ocular biomarkers, including the corneal-related parameters of CCT (0.54 ± 0.04 mm and 0.56 ± 0.03 mm, t = -2.116, P < 0.05), anterior surface CCR (7.53 ± 0.33 mm and 7.84 ± 0.30 mm, t = -5.063, P < 0.05), posterior surface CCR (6.75 ± 0.34 mm and 7.03 ± 0.24 mm, t = -4.864, P < 0.05); as well as those related to anterior chamber depth (ACD) were 3.24 ± 0.26 mm and 3.64 ± 0.26 mm, respectively (t = -8.065, P < 0.05), lens-related parameters (LT) were 3.80 ± 0.19 mm and 3.45 ± 0.16 mm, respectively (t = 10.514, P < 0.05); anterior lens surface curvature radius were 10.02 ± 0.93 mm and 10.52 ± 0.85 mm, respectively (t = -2.962, P < 0.05); posterior lens surface curvature radius were 5.55 ± 0.51 mm and 5.80 ± 0.36 mm, respectively (t = -2.917, P < 0.05), and ocular axis (AL) were 22.60 ± 1.42 mm and 23.45 ± 1.23 mm, respectively (t = -3.332, P < 0.05). Moreover, comparison of refractive status and BCVA between two groups of children showed an SE of -1.23 ± 3.38 D and -0.07 ± 2.00 D (t = -2.206, P < 0.05) and LogMAR (BCVA) of 0.12 ± 0.13 and 0.05 ± 0.11 (t = 3.070, P < 0.05). Analysis of the correlations between age and ocular biomarkers and refractive status of children in the laser and control groups showed correlations between age and ocular biomarkers in the two groups, in which age in the laser group was positively correlated with AL (r = 0.625, P < 0.05) but not with other biomarkers (P > 0.05). Age in the control group was negatively correlated with CCT, ACD, and AL (r = 0.303, 0.468, 0.703, P < 0.05), as well as with LT (r = -0.555, P < 0.05), with no correlation with other biomarkers (P > 0.05). Analysis of the correlation between age and refractive status of children in both groups showed that the age of children in both laser and control groups was negatively correlated with SE (r = -0.528, -0.655, P < 0.05) and LogMAR (BCVA) (r = -0.538, -0.542, P < 0.05). Analysis of the correlations between refractive status and ocular biomarkers in children in the laser and control groups showed that the refractive status in children in the laser group was negatively correlated with AL (r = -0.773, P < 0.05) but not with other biomarkers in this group (P > 0.05). The refractive status of children in the control group was negatively correlated with ACD and AL (r = -0.469, -0.734, P < 0.05), positively correlated with LT (r = 0.364, P < 0.05), and was not correlated with other biomarkers in this group (P > 0.05). Analysis of the correlations of gestational age at birth with ocular biomarkers and refractive status in children in the laser group showed a positive correlation between gestational age at birth and AL (r = 0.435, P < 0.05) but no other correlations with the other biomarkers (P > 0.05). Moreover, gestational age at birth was negatively correlated with SE (r = -0.334, P < 0.05) and LogMAR (BCVA) (r = -0.307, P < 0.05) in children in the laser group. CONCLUSIONS: Compared to full-term infants, the development of CCT, ACD, LT, and AL was relatively delayed after ROP laser surgery, resulting in thin central corneal thickness, steep corneas, shallow anterior chambers, thicker lenses, "rounder" lens morphology, increased refractive power, and short eye axes, leading to the development of myopia. The changes in refractive status were mainly influenced by increased lens thickness. The results of this study showed that the lower the gestational age at birth, the greater the effects on emmetropization in children after ROP, and the more likely the development of myopia.

COMPARISON OF INTRAVITREAL INJECTION OF RANIBIZUMAB VERSUS LASER THERAPY FOR ZONE II TREATMENT-REQUIRING RETINOPATHY OF PREMATURITY.

PURPOSE: To compare the efficacy of intravitreal injection of ranibizumab (IVR) monotherapy and laser therapy for treatment-requiring retinopathy of prematurity (ROP) in Zone II. METHODS: A prospective, randomized, controlled single-center trial was applied from January 2014 to December 2014; infants who were diagnosed as Zone II treatment-requiring ROP (i.e., Zone II Stage 2 or 3 ROP with plus disease) were randomly assigned to receive IVR monotherapy or laser therapy, and the follow-up interval was at least 6 months. Any eyes that developed recurrence of ROP underwent crossover re-treatment. RESULTS: A total of 100 eyes of 50 ethnic Han Chinese infants were enrolled. At the last follow-up, 26 eyes of 13 infants developed recurrence of ROP in the IVR group and 2 eyes of 1 infant developed recurrence of ROP in the laser therapy group. There was a significant statistical difference in the rate of ROP recurrence between IVR and laser therapy to treat Zone II treatment-requiring ROP (P = 0.001). CONCLUSION: Although IVR appears to regress ROP to certain levels and continue to promote the vascularization of peripheral retinal vessels, a substantial proportion of infants developed recurrence of ROP after a single-dose IVR. Therefore, IVR is not recommended as a single-dose monotherapy for Zone II treatment-requiring ROP.

Ultrasound Microbubbles Enhance the Neuroprotective Effect of Mouse Nerve Growth Factor on Intraocular Hypertension-Induced Neuroretina Damage in Rabbits.

Ultrasound microbubble combined optic protection drugs have obvious protective effect on optic nerve damage. This way of targeting drug delivery is becoming more simple, not through the whole body metabolism, avoiding drug via blood circulation when facing the decomposition and the environment in the interference and destruction process of drugs, to maximize the guarantee to reach target organs of drug concentration and to reache the maximum therapeutic effect. The technique of ultrasound microbubbles is safe, controllable, nonimmunogenic, and repeatable. It provides us with a novel idea in the administration of neuroprotective drugs.

[The clinical study on intravitreous injection of ranibizumab for aggressive posterior retinopathy of prematurity].

OBJECTIVE: To observe the retinal vascular development and changes on aggressive posterior retinopathy of prematurity (AP-ROP) by intravitreal ranibizumab, evaluate the therapeutic effect, and provide the basis for clinical treatment. METHODS: The total of 38 eyes of 19 premature infants who were diagnosed as AP-ROP from January 2012 to October 2013 in our hospital were performed intravitreal injection of ranibizumab (0.04 ml). It was observed about the regression of plus diseases, ridges, neovascularization on the ridge and the development of peripheral retinal vessel. Laser photocoagulation were performed for 14 eyes of 7 cases whose plus diseases, ridges and neovascularization on the ridge didn't regress completely after intralvitreal injection of ranibizumab. RESULTS: All infants were found retinopathy regressed and retinal vessels continued to develop peripherally to some degree. Of all infants, 24 eyes of 12 infants were found complete regression of retinopathy, resolution of neovascularization and bleeding and that retinal vessels continued to develop to ora serrata or scarification of peripheral retinopathy. Fourteen eyes of 7 infants were found retinopathy didn't regressed completely and regressed completely after combing intravitreal ranibizumab injection. All 19 infants didn't occure infection, ocular or systemic complications. CONCLUSIONS: The efficacy of intravitreal injection of ranibizumab is good for AP-ROP. It can made ridge, neovascularization on the ridge and plus disease regress, as well as let the retinal vessel continue development. Infants with no regressed retinopathy may need combined laser photocoagulation.

[Membrane peeling combined with intravitreal injection of bevacizumab for treatment of macular epiretinal membrane: analysis of 33 cases].

OBJECTIVE: To assess efficacy of membrane peeling combined with intravitreal injection of bevacizumab in the treatment of macular epiretinal membrane. METHODS: From January, 2012 to June, 2013, 33 patients (33 eyes) with the diagnosis of macular epiretinal membrane underwent vitreous surgery and membrane peeling. The patients were randomly divided into intravitreal bevacizumab group (IVB group) and non-intravitreal bevacizumab group (non-IVB group). All the patients underwent standard three-port vitrectomy and peeling of epiretinal membrane, with intravitreal injection of 1.5 mg bevacizumab at the end of operation in IVB group. The best corrected visual acuity and optical coherence tomography (OCT) were examined before and after the treatment. The patients were followed up for 3-14 months (mean 6.5 months). RESULTS: Macular epiretinal membranes were successfully peeled during operation in all the patients without postoperative intraocular infection or bleeding. Fifteen eyes received vitrectomy combined with intravitreal injection of bevacizumab, and 18 underwent only vitreous operation and membrane peeling. At the end of the follow up, the visual acuity improved in 11 eyes (73.3%) in IVB group, as compared to 13 eyes (72.2%) in the non-IVB group (P=0.627). Central macular thickness decreased by 143∓62 µm in IVB group and by 96∓28 µm in non-IVB group, showing a significant difference between the two groups (t=5.564, P<0.01). CONCLUSION: Vitrectomy and membrane peeling combined with intravitreal injection of bevacizumab can promote the recovery of macular morphology but not visual function, and its clinical use still needs to be tested in a long-term and large-sample randomized controlled study.