Development and validation of a deep learning system to classify aetiology and predict anatomical outcomes of macular hole.

AIMS: To develop a deep learning (DL) model for automatic classification of macular hole (MH) aetiology (idiopathic or secondary), and a multimodal deep fusion network (MDFN) model for reliable prediction of MH status (closed or open) at 1 month after vitrectomy and internal limiting membrane peeling (VILMP). METHODS: In this multicentre retrospective cohort study, a total of 330 MH eyes with 1082 optical coherence tomography (OCT) images and 3300 clinical data enrolled from four ophthalmic centres were used to train, validate and externally test the DL and MDFN models. 266 eyes from three centres were randomly split by eye-level into a training set (80%) and a validation set (20%). In the external testing dataset, 64 eyes were included from the remaining centre. All eyes underwent macular OCT scanning at baseline and 1 month after VILMP. The area under the receiver operated characteristic curve (AUC), accuracy, specificity and sensitivity were used to evaluate the performance of the models. RESULTS: In the external testing set, the AUC, accuracy, specificity and sensitivity of the MH aetiology classification model were 0.965, 0.950, 0.870 and 0.938, respectively; the AUC, accuracy, specificity and sensitivity of the postoperative MH status prediction model were 0.904, 0.825, 0.977 and 0.766, respectively; the AUC, accuracy, specificity and sensitivity of the postoperative idiopathic MH status prediction model were 0.947, 0.875, 0.815 and 0.979, respectively. CONCLUSION: Our DL-based models can accurately classify the MH aetiology and predict the MH status after VILMP. These models would help ophthalmologists in diagnosis and surgical planning of MH.

Corrigendum: Risk factors of falls in elderly patients with visual impairment.

[This corrects the article DOI: 10.3389/fpubh.2022.984199.].

Risk factors of falls in elderly patients with visual impairment.

Objective: To examine the risk factors for falls in elderly patients with visual impairment (VI) and assess the predictive performance of these factors. Methods: Between January 2019 and March 2021, a total of 251 elderly patients aged 65-92 years with VI were enrolled and then prospectively followed up for 12 months to evaluate outcomes of accidental falls via telephone interviews. Information of demographics and lifestyle, gait and balance deficits, and ophthalmic and systemic conditions were collected during baseline visits. Forward stepwise multivariable logistic regression analysis was performed to identify independent risk factors of falls in elderly patients with VI, and a derived nomogram was constructed. Results: A total of 143 falls were reported in 251 elderly patients during follow-up, with an incidence of 56.97%. The risk factors for falls in elderly patients with VI identified by multivariable logistic regression were women [odds ratio (OR), 95% confidence interval (CI): 2.71, 1.40-5.27], smoking (3.57, 1.34-9.48), outdoor activities/3 months (1.31, 1.08-1.59), waking up frequently during the night (2.08, 1.15-3.79), disorders of balance and gait (2.60, 1.29-5.24), glaucoma (3.12, 1.15-8.44), other retinal degenerations (3.31, 1.16-9.43) and best-corrected visual acuity (BCVA) of the better eye (1.79, 1.10-2.91). A nomogram was developed based on the abovementioned multivariate analysis results. The area under receiver operating characteristic curve of the predictive model was 0.779. Conclusions: Gender, smoking, outdoor activities, waking up at night, disorders of balance and gait, glaucoma, other retinal degeneration and BCVA of the better eye were independent risk factors for falls in elderly patients with VI. The predictive model and derived nomogram achieved a satisfying prediction of fall risk in these individuals.

Causes and Risk Factors of Repeated Hospitalization among Patients with Diabetic Retinopathy.

Purpose: To identify the causes and risk factors of repeated hospitalization among patients with diabetic retinopathy (DR). Methods: Our study retrospectively examined the data of DR patients who were readmitted for treatments to the Department of Ophthalmology, Guangdong Provincial People's Hospital between January 2012 and July 2021. We first analyzed the main causes of repeated admissions and then divided the patients into three groups according to the times of readmissions. Ordinal logistic regression was performed to determine the impact of patients' demographic and clinical characteristics. Moreover, comparisons of the length of stay and the hospitalization cost of DR patients with repeated admission causes were conducted. Results: Among 2592 hospital discharges of 827 patients who experienced at least two hospitalizations, the major causes of repeated hospitalization were macular edema (30.83%), vitreous hemorrhage (29.09%), cataract (22.76%), proliferative membrane formation (6.91%), silicone oil removal (4.71%), retinal detachment (4.44%), and glaucoma (4.17%). The results of ordinal logistic regression showed that younger patients with medical insurance and local residence have a higher risk of repeated hospitalization (p < 0.05). Furthermore, patients readmitted for vitreous hemorrhage, proliferative membrane formation, and retinal detachment experienced longer length of hospital stay and higher hospitalization cost (p < 0.001). Conclusions: Multiple causes and risk factors contribute to repeated hospitalization, imposing a substantial physical and economic burden on DR patients. A better understanding of these causes and risk factors of readmission may lead to lowering such risks and alleviating patients' burden.

Development and Validation of a Deep Learning Model to Predict the Occurrence and Severity of Retinopathy of Prematurity.

Importance: Retinopathy of prematurity (ROP) is the leading cause of childhood blindness worldwide. Prediction of ROP before onset holds great promise for reducing the risk of blindness. Objective: To develop and validate a deep learning (DL) system to predict the occurrence and severity of ROP before 45 weeks' postmenstrual age. Design, Setting, and Participants: This retrospective prognostic study included 7033 retinal photographs of 725 infants in the training set and 763 retinal photographs of 90 infants in the external validation set, along with 46 characteristics for each infant. All images of both eyes from the same infant taken at the first screening were labeled according to the final diagnosis made between the first screening and 45 weeks' postmenstrual age. The DL system was developed using retinal photographs from the first ROP screening and clinical characteristics before or at the first screening in infants born between June 3, 2017, and August 28, 2019. Exposures: Two models were specifically designed for predictions of the occurrence (occurrence network [OC-Net]) and severity (severity network [SE-Net]) of ROP. Five-fold cross-validation was applied for internal validation. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity to evaluate the performance in ROP prediction. Results: This study included 815 infants (450 [55.2%] boys) with mean birth weight of 1.91 kg (95% CI, 1.87-1.95 kg) and mean gestational age of 33.1 weeks (95% CI, 32.9-33.3 weeks). In internal validation, mean AUC, accuracy, sensitivity, and specificity were 0.90 (95% CI, 0.88-0.92), 52.8% (95% CI, 49.2%-56.4%), 100% (95% CI, 97.4%-100%), and 37.8% (95% CI, 33.7%-42.1%), respectively, for OC-Net to predict ROP occurrence and 0.87 (95% CI, 0.82-0.91), 68.0% (95% CI, 61.2%-74.8%), 100% (95% CI, 93.2%-100%), and 46.6% (95% CI, 37.3%-56.0%), respectively, for SE-Net to predict severe ROP. In external validation, the AUC, accuracy, sensitivity, and specificity were 0.94, 33.3%, 100%, and 7.5%, respectively, for OC-Net, and 0.88, 56.0%, 100%, and 35.3%, respectively, for SE-Net. Conclusions and Relevance: In this study, the DL system achieved promising accuracy in ROP prediction. This DL system is potentially useful in identifying infants with high risk of developing ROP.

Effect of High Myopia and Cataract Surgery on the Correlation Between Diabetic Retinopathy and Chronic Kidney Disease.

Purpose: To investigate the effect of high myopia and cataract surgery on the grading of diabetic retinopathy (DR) and their roles in the correlation between DR and chronic kidney disease (CKD). Methods: A total of 1,063 eyes of 1,063 diabetic patients were enrolled. We conducted binary and multiple multivariate regressions to analyze the ocular and systemic risk factors of DR. Based on the presence of myopia and history of cataract surgery, we divided the cases into four subgroups, namely those with high myopia, with the history of cataract surgery, with both conditions, and with neither, then determined the correlation between the stages of DR and CKD in each subgroup. Results: In the binary analysis, high myopia was identified as the protective factor for DR odds ratio (OR): 0.312 [95% confidence interval (CI): 0.195-0.500, p < 0.001], whereas cataract surgery was one of the independent risk factors for DR [OR: 2.818 (95% CI: 1.507-5.273), p = 0.001]. With increased stages of DR, high myopia played an increasingly protective role [mild non-proliferative DR (NPDR), OR = 0.461, p = 0.004; moderate NPDR OR = 0.217, p = 0.003; severe NPDR, OR = 0.221, p = 0.008; proliferative DR (PDR), OR = 0.125, p = 0.001], whereas cataract surgery became a stronger risk factor, especially in PDR (mild NPDR, OR = 1.595, p = 0.259; moderate NPDR, OR = 3.955, p = 0.005; severe NPDR, OR = 6.836, p < 0.001; PDR, OR = 9.756, p < 0.001). The correlation between the stages of DR and CKD in the group with neither high myopia nor cataract surgery history was the highest among all subgroups. Conclusion: High myopia was a protective factor, whereas cataract surgery is a risk factor for DR, and both factors showed stronger effects throughout the (natural disease) grading of DR. The stages of DR and CKD showed a higher correlation after adjustment of the ocular confounding factors.

Retinal Neurovascular Impairment in Non-diabetic and Non-dialytic Chronic Kidney Disease Patients.

Background: Widespread neural and microvascular injuries are common in chronic kidney disease (CKD), increasing risks of neurovascular complications and mortality. Early detection of such changes helps assess the risks of neurovascular complications for CKD patients. As an extension of central nervous system, the retina provides a characteristic window to observe neurovascular alterations in CKD. This study aimed to determine the presence of retinal neurovascular impairment in different stages of CKD. Methods: One hundred fifteen non-diabetic and non-dialytic CKD patients of all stages and a control group of 35 healthy subjects were included. Retinal neural and microvascular parameters were obtained by optical coherence tomography angiography (OCTA) examination. Results: CKD 1-2 group (versus control group) had greater odds of having decreased retinal ganglion cell-inner plexiform layer thickness (GC-IPLt) (odds ratio [OR]: 0.92; 95% confidence interval [CI]: 0.86-0.98), increased ganglion cell complex-focal loss volume (GCC-FLV) (OR: 3.51; 95% CI: 1.27-9.67), and GCC-global loss volume (GCC-GLV) (OR: 2.48; 95% CI: 1.27-4.82). The presence of advanced stages of CKD (CKD 3-5 group versus CKD 1-2 group) had greater odds of having decreased retinal vessel density in superficial vascular plexus (SVP)-WholeImage (OR: 0.77, 95% CI: 0.63-0.92), SVP-ParaFovea (OR: 0.83, 95% CI: 0.71-0.97), SVP-ParaFovea (OR: 0.76, 95% CI: 0.63-0.91), deep vascular plexus (DVP)-WholeImage (OR: 0.89, 95% CI: 0.81-0.98), DVP-ParaFovea (OR: 0.88, 95% CI: 0.78-0.99), and DVP-PeriFovea (OR: 0.90, 95% CI: 0.83-0.98). Besides, stepwise multivariate linear regression among CKD patients showed that ß2-microglobulin was negatively associated with GC-IPLt (ß: -0.294; 95% CI: -0.469 ∼ -0.118), and parathyroid hormone was positively associated with increased GCC-FLV (ß: 0.004; 95% CI: 0.002∼0.006) and GCC-GLV (ß: 0.007; 95% CI: 0.004∼0.01). Urine protein to creatinine ratio was positively associated with increased GCC-FLV (ß: 0.003; 95% CI: 0.001∼0.004) and GCC-GLV (ß: 0.003; 95% CI: 0.001∼0.006). Conclusion: Retinal neuronal impairment is present in early stages of CKD (stages 1-2), and it is associated with accumulation of uremic toxins and higher UACR, while retinal microvascular hypoperfusion, which is associated with worse eGFR, was only observed in relatively advanced stages of CKD (stages 3-5). The results highlight the importance of monitoring retinal neurovascular impairment in different stages of CKD.

Machine learning-based prediction of anatomical outcome after idiopathic macular hole surgery.

BACKGROUND: To develop a machine learning (ML) model for the prediction of the idiopathic macular hole (IMH) status at 1 month after vitrectomy and internal limiting membrane peeling (VILMP) surgery. METHODS: A total of 288 IMH eyes from four ophthalmic centers were enrolled. All eyes underwent optical coherence tomography (OCT) examinations upon admission and one month after VILMP. First, 1,792 preoperative macular OCT parameters and 768 clinical variables of 256 eyes from two ophthalmic centers were used to train and internally validate ML models. Second, 224 preoperative macular OCT parameters and 96 clinical variables of 32 eyes from the other two centers were utilized for external validation. To fulfill the purpose of predicting postoperative IMH status (i.e., closed or open), five ML algorithms were trained and internally validated by the ten-fold cross-validation method, while the best-performing algorithm was further tested by an external validation set. RESULTS: In the internal validation, the mean area under the receiver operating characteristic curves (AUCs) of the five ML algorithms were 0.882-0.951. The AUC, accuracy, sensitivity, and specificity of the best-performing algorithm (i.e., random forest, RF) were 0.951, 0.892, 0.973, and 0.904, respectively. In the external validation, the AUC of RF was 0.940, with an accuracy of 0.875, a specificity of 0.875, and a sensitivity of 0.958. CONCLUSIONS: Based on the preoperative OCT parameters and clinical variables, our ML model achieved remarkable accuracy in predicting IMH status after VILMP. Therefore, ML models may help optimize surgical planning for IMH patients in the future.

Reduced Retinal Microvascular Perfusion in Patients With Stroke Detected by Optical Coherence Tomography Angiography.

Currently there is a shortage of biomarkers for stroke, one of the leading causes of death and disability in aging populations. Retinal vessels offer a unique and accessible "window" to study the microvasculature in vivo. However, the relationship between the retinal microvasculature and stroke is not entirely clear. To investigate the retinal microvascular characteristics in stroke, we recruited patients with stroke and age-matched control subjects from a tertiary hospital in China. The macular vessel density (VD) in the superficial capillary plexus (SCP) and deep capillary plexus (DCP), foveal avascular zone (FAZ) metrics, and optical coherence tomography angiography (OCTA) measured optic disc VD were recorded for analysis. A total of 189 patients with stroke and 195 control subjects were included. After adjusting for sex, visual acuity, systolic and diastolic blood pressure, a history of smoking, levels of hemoglobulin (HbA1c), cholesterol, and high-density lipoprotein (HDL), the macular VD of SCP and DCP in all sectors was decreased in patients with stroke. In the stroke group, the VD around the FAZ and the VD of the optic disk were lower. Logistic regression found the parafovea-superior-hemi VD of DCP > 54.53% [odds ratio (OR): 0.169] as a protective factor of stroke. Using the integration of all OCTA parameters and traditional risk factors, the area under the receiver operating characteristic (AUC) curve of distinguishing patients with stroke was 0.962, with a sensitivity of 0.944 and a specificity of 0.871. Our study demonstrates that the retinal VD is decreased in patients with stroke independently of the traditional risk factors of stroke, which may shed light on the monitoring of stroke using the retinal microvascular parameters.

Factors Associated With the Presence of Foveal Bulge in Eyes With Resolved Diabetic Macular Edema.

Purpose: To evaluate factors associated with the presence of foveal bulge (FB) in resolved diabetic macular edema (DME) eyes. Methods: A total of 165 eyes with complete integrity of ellipsoid zone (EZ) at the fovea and resolved DME were divided into two groups according to the presence of FB at 6 months after intravitreal injection of ranibizumab treatment. Best-corrected visual acuity (BCVA), central foveal thickness (CFT), outer nuclear layer (ONL) thickness, height of serous retinal detachment (SRD) and non-SRD, and inner segment (IS) and outer segment (OS) lengths of the two groups were measured and compared at baseline and each follow-up. The correlations between the presence of FB and pre- and post-treatment factors were determined by logistic regression analysis. Results: At baseline, BCVA was significantly better, and CFT and incidence and height of SRD were significantly lower in the FB (+) group (all P < 0.05). At 6 months, FB was present in 65 (39.39%) eyes. Post-treatment BCVA was significantly better and OS length was significantly longer in the FB (+) group at 6 months (all P < 0.05). Multivariate analysis identified younger age, better BCVA, and lower CFT before treatment as significant predictors of the existence of FB at 6 months (all P < 0.05). At 6 months, better BCVA and longer OS length were significantly correlated with the existence of FB (all P < 0.05). Conclusions: Factors associated with the presence of FB after the resolution of DME include younger age, better baseline BCVA and lower baseline CFT, and better post-treatment BCVA and longer post-treatment OS length.

Investigation of Correlations Between Optical Coherence Tomography Biomarkers and Visual Acuity in X-Linked Retinoschisis.

PURPOSE: To investigate the imaging biomarkers of spectral-domain optical coherence tomography (SD-OCT) and their correlations with age and best-corrected visual acuity (BCVA) in patients with X-linked retinoschisis (XLRS). METHODS: OCT images of 72 eyes of 39 patients with confirmed XLRS were obtained to assess imaging biomarkers, including but not limited to the automatic evaluation of foveal thickness, central subfield thickness (CST), macular volume, and the manual measurement of area of macular schisis cavity (AMS). Correlations between age/BCVA and all OCT parameters were computed as well. RESULTS: In this study, median age was 10.5 (8-12) years old and median BCVA was 0.90 (0.70-1.00) logarithm of the minimum angle of resolution. Macular retinoschisis was found in all affected eyes, with peripheral retinoschisis (PRS) in 34 (47.2%) eyes. Cystic cavities most frequently affected inner nuclear layer (100%) in the macula. Ellipsoid zone (EZ) defects occurred in 53 (73.6%) eyes. As for correlation, BCVA was significantly correlated with several OCT parameters, including CST, AMS, EZ defect, PRS and vitreomacular adhesion, whereas no correlation was found between age and any OCT parameter. CONCLUSION: Explicable OCT imaging biomarkers such as CST, AMS, and photoreceptor defects were identified and may serve as reference parameters or potential regions of interest for future observational and interventional research design and result interpretation.