Artificial intelligence for the diagnosis of retinopathy of prematurity: A systematic review of current algorithms

BACKGROUND/OBJECTIVES

With the increasing survival of premature infants, there is an increased demand to provide adequate retinopathy of prematurity (ROP) services. Wide field retinal imaging (WFDRI) and artificial intelligence (AI) have shown promise in the field of ROP and have the potential to improve the diagnostic performance and reduce the workload for screening ophthalmologists. The aim of this review is to systematically review and provide a summary of the diagnostic characteristics of existing deep learning algorithms.

SUBJECT/METHODS

Two authors independently searched the literature, and studies using a deep learning system from retinal imaging were included. Data were extracted, assessed and reported using PRISMA guidelines.

RESULTS

Twenty-seven studies were included in this review. Nineteen studies used AI systems to diagnose ROP, classify the staging of ROP, diagnose the presence of pre-plus or plus disease, or assess the quality of retinal images. The included studies reported a sensitivity of 71%-100%, specificity of 74-99% and area under the curve of 91-99% for the primary outcome of the study. AI techniques were comparable to the assessment of ophthalmologists in terms of overall accuracy and sensitivity. Eight studies evaluated vascular severity scores and were able to accurately differentiate severity using an automated classification score.

CONCLUSION

Artificial intelligence for ROP diagnosis is a growing field, and many potential utilities have already been identified, including the presence of plus disease, staging of disease and a new automated severity score. AI has a role as an adjunct to clinical assessment; however, there is insufficient evidence to support its use as a sole diagnostic tool currently.

Retinopathy of prematurity in India - what can we learn from the polio legacy?

Retinopathy of prematurity (ROP) is a vasoproliferative disease of the preterm retina that has the potential to cause vision impairment and blindness. Timely screening and treatment are hence critical for infants at risk for ROP. Screening for ROP is challenging in India owing to the limited resources, a vast at-risk population and lack of awareness among paediatricians and the public. Addressing ROP in India requires a comprehensive approach involving multiple sectors, considering the magnitude of the problem and the expected increase in need for ROP services due to the increased survival of preterm infants following improvements in neonatal care. The success of the Global Polio Eradication Initiative (GPEI) offers valuable lessons for improving ROP services in developing nations by applying its strategies. An approach for primary and secondary prevention of ROP is proposed, and the current challenges and a neonatal-led care model for ROP are discussed.

Retinopathy of prematurity and placental histopathology findings: A retrospective cohort study

Aim

Retinopathy of prematurity (ROP) is a biphasic vaso-proliferative disease that has the potential to cause blindness. In addition to prematurity and hyperoxia, perinatal infection and inflammation have been reported to play a critical role in the pathogenesis of ROP. The aim of this study was to assess the association between placental inflammation and the severity of ROP.

Methods

A retrospective study of infants (<30 weeks of gestational age) born at the King Edward Memorial Hospital, a tertiary perinatal center in Western Australia.

Results

A total of 878 infants were included in this study (ROP stage 0-2 = 829; 3 or more = 49). The presence of maternal chorioamnionitis appeared to show signs of an association with reduced odds of severe ROP: mild chorioamnionitis OR=0.43 (95% CI: 0.17, 1.05) and severe chorioamnionitis OR=0.68 (95% CI: 0.29, 1.60). A strong association was observed for oxygen supplementation at 36 weeks (OR: 5.16; p < 0.001), exposure to postnatal steroids (OR: 6.65; p < 0.001), and receipt of platelet transfusion (OR: 8.21; p < 0.001).

Conclusion

Maternal chorioamnionitis or fetal chorioamnionitis was associated with reduced odds of severe ROP. A strong association was found in infants who needed oxygen supplementation at 36 weeks and those who required steroids or platelets in the postnatal period.

Comparison of wide field imaging by nurses with indirect ophthalmoscopy by ophthalmologists for retinopathy of prematurity: a diagnostic accuracy study

OBJECTIVES

Retinopathy of prematurity (ROP) is a vasoproliferative disease of the preterm retina with the potential to cause irreversible blindness. Timely screening and treatment of ROP are critical. Neonatal nurses trained in wide field digital retinal photography (WFDRP) for screening may provide a safe and effective strategy to reduce the burden of ophthalmologists in performing binocular indirect ophthalmoscopy (BIO). The objective of the study was to determine the diagnostic accuracy of WFDRP in the diagnosis of referral warranting ROP (RWROP).

DESIGN

Prospective diagnostic accuracy study.

SETTING

A tertiary neonatal intensive care unit in Perth, Western Australia.

PARTICIPANTS

Preterm infants who fulfilled the Australian ROP screening criteria (gestational age (GA) <31 weeks, birth weight (BW) <1250 g).

INTERVENTION

Sets of 5-6 images per eye (index test) were obtained within 24-48 hours prior to or after the BIO (reference standard), and uploaded onto a secured server. A wide field digital camera (RetCam, Natus, Pleasanton, California, USA) was used for imaging. A paediatric ophthalmologist performed the BIO. The ophthalmologists performing BIO versus reporting the images were masked to each other's findings.

PRIMARY OUTCOME

The area under the receiver operating characteristic (ROC) curve was used as a measure of accuracy of WFDRP to diagnose RWROP.

RESULTS

A total of 85 infants (mean BW; 973.43 g, mean GA; 29 weeks) underwent a median of two sessions of WFDRP. There were 188 episodes of screening with an average of five images per eye. WFDRP identified RWROP in 7.4% (14/188 sessions) of examinations. In one infant, BIO showed bilateral plus disease and WFDRP did not pick up the plus disease. WFDRP image interpretation had a sensitivity of 80%, specificity of 94.5% for the detection of RWROP. The 'area under the ROC curve' was 88% when adjusted for covariates.

CONCLUSIONS

WFDRP by neonatal nurses was feasible and effective for diagnosing RWROP in our set up.

TRIAL REGISTRATION NUMBER

ACTRN12616001386426.