Development and international validation of custom-engineered and code-free deep-learning models for detection of plus disease in retinopathy of prematurity: a retrospective study.

BACKGROUND: Retinopathy of prematurity (ROP), a leading cause of childhood blindness, is diagnosed through interval screening by paediatric ophthalmologists. However, improved survival of premature neonates coupled with a scarcity of available experts has raised concerns about the sustainability of this approach. We aimed to develop bespoke and code-free deep learning-based classifiers for plus disease, a hallmark of ROP, in an ethnically diverse population in London, UK, and externally validate them in ethnically, geographically, and socioeconomically diverse populations in four countries and three continents. Code-free deep learning is not reliant on the availability of expertly trained data scientists, thus being of particular potential benefit for low resource health-care settings. METHODS: This retrospective cohort study used retinal images from 1370 neonates admitted to a neonatal unit at Homerton University Hospital NHS Foundation Trust, London, UK, between 2008 and 2018. Images were acquired using a Retcam Version 2 device (Natus Medical, Pleasanton, CA, USA) on all babies who were either born at less than 32 weeks gestational age or had a birthweight of less than 1501 g. Each images was graded by two junior ophthalmologists with disagreements adjudicated by a senior paediatric ophthalmologist. Bespoke and code-free deep learning models (CFDL) were developed for the discrimination of healthy, pre-plus disease, and plus disease. Performance was assessed internally on 200 images with the majority vote of three senior paediatric ophthalmologists as the reference standard. External validation was on 338 retinal images from four separate datasets from the USA, Brazil, and Egypt with images derived from Retcam and the 3nethra neo device (Forus Health, Bengaluru, India). FINDINGS: Of the 7414 retinal images in the original dataset, 6141 images were used in the final development dataset. For the discrimination of healthy versus pre-plus or plus disease, the bespoke model had an area under the curve (AUC) of 0·986 (95% CI 0·973-0·996) and the CFDL model had an AUC of 0·989 (0·979-0·997) on the internal test set. Both models generalised well to external validation test sets acquired using the Retcam for discriminating healthy from pre-plus or plus disease (bespoke range was 0·975-1·000 and CFDL range was 0·969-0·995). The CFDL model was inferior to the bespoke model on discriminating pre-plus disease from healthy or plus disease in the USA dataset (CFDL 0·808 [95% CI 0·671-0·909, bespoke 0·942 [0·892-0·982]], p=0·0070). Performance also reduced when tested on the 3nethra neo imaging device (CFDL 0·865 [0·742-0·965] and bespoke 0·891 [0·783-0·977]). INTERPRETATION: Both bespoke and CFDL models conferred similar performance to senior paediatric ophthalmologists for discriminating healthy retinal images from ones with features of pre-plus or plus disease; however, CFDL models might generalise less well when considering minority classes. Care should be taken when testing on data acquired using alternative imaging devices from that used for the development dataset. Our study justifies further validation of plus disease classifiers in ROP screening and supports a potential role for code-free approaches to help prevent blindness in vulnerable neonates. FUNDING: National Institute for Health Research Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and the University College London Institute of Ophthalmology. TRANSLATIONS: For the Portuguese and Arabic translations of the abstract see Supplementary Materials section.

Screening and Risk Factors for Retinopathy of Prematurity in a Tertiary Care Hospital in Cairo, Egypt.

Purpose: To evaluate the retinopathy of prematurity (ROP) prevalence, risk factors and screening outcome in a tertiary hospital in Cairo, Egypt. Methods: A prospective observational study was done in Neonatal Intensive Care Unit in Ain Shams University Hospital. A total of 159 premature infants were screened for ROP based on the most inclusive criteria reported to date. Screening included premature infants with gestational age (GA) of ≤34 weeks or birth weight (BW) of ≤2000 grams, or GA >34 weeks or BW >2000 grams, with multiple co-morbidities. The prevalence of ROP, plus disease and their correlation with risk factors of interest were studied. Results: The GA of the included infants ranged from 27 to 36 weeks, mean (SD) 31.87 (± 1.81) weeks. The BW ranged from 640 to 3900 grams, mean (SD) 1784.71 (± 560.30) grams. The prevalence of ROP more than stage 0 was 25.8% (41 infants), 7.3% of the cases (11 infants) showed plus disease and 6.3% (10 infants) showed severe ROP requiring treatment. Of those, 2 cases (20%) fell outside the British Guideline's criteria for Screening. There was a highly significant (p < 0.0001) correlation between ROP more than stage 0 and low GA, low BW, mechanical ventilation, respiratory distress syndrome, necrotizing enterocolitis, intraventricular haemorrhage, and blood transfusion. No significant correlation was found between appearance of ROP more than stage 0 and gender (p = 0.911), patent ductus arteriosus (p =0.187), or sepsis (p =0.998). Conclusion: ROP is a significant problem in the premature infants in Egypt. Extremely premature infants with lower BW are more prone to develop ROP. However, cases with higher GA and BW than mentioned in the British guidelines screening criteria especially with multiple comorbidities showed severe ROP requiring intervention, which implies the need to develop a screening guideline for the Egyptian population.

Analysis of a two-year independent screening effort for retinopathy of prematurity in rural Egypt.

BACKGROUND: The third epidemic of retinopathy of prematurity (ROP) has majorly involved middle income countries in which tailored screening and local guidelines require development. The data regarding ROP prevalence and cutoff numbers for screening in Egypt are lacking. METHODS: Retrospective analysis of an independent screening effort spanning 2 years (February 2019 to February 2021) and involving 32 neonatal care units within Sharkia governorate, Egypt. Infants of gestational age (GA) ≤ 34 weeks and/or birth weight (BW) ≤ 2000 g were included, as well as those with unstable clinical course. Two eyecare centers located in Sharkia and Cairo governorates served as referral centers for any required interventions. RESULTS: Of the 276 screened infants, 133 (48.2%) had some form of ROP that was bilateral in 127 (95.5%) of them. Aggressive posterior ROP (AP-ROP) was detected in both eyes of 24 infants (8.7%). The median (IQR) GA of infants with ROP was 32 (30-34) weeks, and the median (IQR) BW was 1600 (1350-2000) g. Sixty-three infants (47.4%) required treatment. Of the total 84 eyes that primarily were treated, 73 (86.9%) received intravitreal ranibizumab, 8 (9.5%) underwent laser ablation therapy, and 3 eyes (3.6%) underwent surgery. Recurrence rate was 16.7% (14 eyes). Final outcome was favorable in 83 eyes (98.8%). Applying the American Academy criteria would have led to the missing of 36.8% of infants with ROP and 28.6% of those requiring treatment in our sample. CONCLUSION: The incidence of both ROP and AP-ROP in the Egyptian rural setting appears to be in the high end of global reported rates. Prevention measures should urgently be planned and implemented.

Experience with Intravitreal Ranibizumab as an Adjunct to Ablation Therapy in Eyes with Exudative Coats' Disease.

BACKGROUND: Coats' disease is a rare entity with retinal vascular telangiectasia that can progress to exudative retinal detachment, neovascular glaucoma, and a blind painful eye requiring enucleation. Despite recent therapeutic advances decreasing the need for enucleation, no consensus exists about the optimum management of exudative Coats' disease. The use of intravitreal anti-vascular endothelial growth factor agents as an adjunct to ablation therapy has been shown to achieve favorable outcomes, but some reports suggest an increased incidence of vitreoretinal (VR) fibrosis and tractional retinal detachment (TRD). METHODS: We retrospectively reviewed records of patients presenting with exudative Coats' disease (stages 2 and 3) from April 2016 till November 2020. Extracted data included clinical and radiological assessment, stage (Shields' classification), interventions, and follow-up. RESULTS: Sixteen eyes were included in the final analysis, of which 4 (25%) were stage 2 and 12 (75%) were stage 3. All eyes underwent intravitreal ranibizumab injection combined with ablation therapy, 14 (87.5%) underwent cryotherapy, 4 (25%) underwent laser ablation, 3 (18.75%) underwent external subretinal fluid drainage, and 3 (18.75%) underwent buckle or vitrectomy surgery. After a median follow-up of 16 months, 11 eyes (68.75%) had complete resolution, 4 (25%) had incomplete resolution, and only one (6.25%) progressed but did not require enucleation. Three eyes (18.75%) developed VR fibrosis, but none progressed to TRD. CONCLUSION: Combining intravitreal ranibizumab injection with ablation therapy is effective in managing exudative Coats' disease. External drainage should be preserved for when ablation therapy is not feasible. Future prospective trials with pre-defined outcomes are required.