Fairness and generalisability in deep learning of retinopathy of prematurity screening algorithms: a literature review.

BACKGROUND: Retinopathy of prematurity (ROP) is a vasoproliferative disease responsible for more than 30 000 blind children worldwide. Its diagnosis and treatment are challenging due to the lack of specialists, divergent diagnostic concordance and variation in classification standards. While artificial intelligence (AI) can address the shortage of professionals and provide more cost-effective management, its development needs fairness, generalisability and bias controls prior to deployment to avoid producing harmful unpredictable results. This review aims to compare AI and ROP study's characteristics, fairness and generalisability efforts. METHODS: Our review yielded 220 articles, of which 18 were included after full-text assessment. The articles were classified into ROP severity grading, plus detection, detecting treatment requiring, ROP prediction and detection of retinal zones. RESULTS: All the article's authors and included patients are from middle-income and high-income countries, with no low-income countries, South America, Australia and Africa Continents representation.Code is available in two articles and in one on request, while data are not available in any article. 88.9% of the studies use the same retinal camera. In two articles, patients' sex was described, but none applied a bias control in their models. CONCLUSION: The reviewed articles included 180 228 images and reported good metrics, but fairness, generalisability and bias control remained limited. Reproducibility is also a critical limitation, with few articles sharing codes and none sharing data. Fair and generalisable ROP and AI studies are needed that include diverse datasets, data and code sharing, collaborative research, and bias control to avoid unpredictable and harmful deployments.

Comparison of the detection rates for retinopathy of prematurity (ROP) of the 2013 Philippine Academy of Ophthalmology (PAO) Revised Philippine Guidelines and the 2005 PAO-Philippine Pediatric Society (PPS) Guidelines for ROP screening in the Philippine General Hospital: a 5-year review.

OBJECTIVES: Compare the detection rates of the 2013 Philippine Academy of Ophthalmology (PAO) guidelines for retinopathy of prematurity (ROP) screening and the 2005 PAO-Philippine Pediatric Society guidelines in identifying infants who develop ROP in the 5-year study period in the Philippine General Hospital (PGH). Secondary objectives include determination of ROP prevalence; correlation of gestational age (GA), birth weight (BW) and other risk factors to ROP; and identification of the most common intervention. METHODS AND ANALYSIS: Retrospective cross-sectional study of ROP records between 1 December 2013 and 30 November 2018 from the Medical Retina Service of the Department of Ophthalmology and Visual Sciences of the institution was studied. Variables with p value <0.05 were considered significant. STATA V.14 was used for all analysis. RESULTS: Only 851 of 898 infants screened for ROP were included in the study. Of these 698 would have been screened based on 2005 guidelines. All 118 infants with ROP were identified by both guidelines. Detection rate was higher using the 2005 than the 2013 guidelines (16.9% vs 13.3%, p value=0.0496). ROP prevalence was 9.7%. Among those with ROP, 70% have at least one identified risk factor, topped by sepsis, pneumonia, hyaline membrane disease, blood transfusion and oxygen supplementation. Only 8% required intervention consisting of laser, anti-vascular endothelial growth factor injection, surgery or in combination. CONCLUSION: In PGH, no infants with ROP were missed using the 2005 recommendations. There was no added benefit of increasing threshold for BW and GA as recommended by the 2013 PAO guidelines. Screening guidelines should, however, be tailored to institutional needs, requirements and experience.