Neonatal retinal hemorrhage and retinopathy of prematurity: a critical review of current practices and future directions

Abstract

Neonatal retinal hemorrhage (RH) and retinopathy of prematurity (ROP) are significant causes of visual impairment in preterm infants, necessitating effective diagnostic and therapeutic strategies. This review synthesizes current approaches to RH and ROP management, highlighting their limitations and the critical need for standardized diagnostic criteria. Indirect ophthalmoscopy, while accessible, suffers from subjective variability, whereas wide-field digital retinal imaging, despite superior visualization, is limited by cost and training requirements. Vaginal delivery and prematurity-related factors, such as fragile retinal vasculature and neonatal complications (e.g., sepsis and hyperbilirubinemia), significantly increase RH and ROP risk, with severe RH correlating with ROP progression. Advances in digital retinal imaging and artificial intelligence (AI) offer promising solutions for early detection, enabling timely interventions such as anti-VEGF therapy to improve visual outcomes. Future directions include developing universal diagnostic standards, exploring novel treatments such as probiotics, and leveraging technologies such as optical coherence tomography. Enhanced screening programs, collaborative research, and public awareness are essential to optimize RH and ROP management, ultimately improving long-term visual prognosis for affected infants.

Keywords

artificial intelligence digital retinal imaging early intervention neonatal retinal hemorrhage retinopathy of prematurity

Background

Emerging evidence has established a correlation between neonatal retinal disorders and subsequent pediatric ophthalmic conditions, including strabismus, amblyopia, nystagmus, and anisometropia.¹ Retinal hemorrhage (RH), recognized as the most frequent retinal pathology in neonates, has gained considerable attention due to the proliferation of neonatal eye disease screening programs. The enhanced detection of RH underscores its significance in the context of visual development and potential long-term sequelae.

Neonatal RH, particularly prevalent among preterm infants, often intersects with critical conditions such as retinopathy of prematurity (ROP).² This intersection demands a comprehensive understanding of RH’s pathophysiology, clinical manifestations, and the extent of its impact on the developing visual system. Early detection and management of RH are paramount, given its potential to cause significant visual impairment if left unaddressed.

This review critically reviews diagnostic and therapeutic strategies for neonatal RH and ROP, addressing their efficacy and limitations to guide clinical advancements. Its necessity stems from persistent gaps in neonatal retinal care, as pre-2023 studies often rely on outdated, high-income country data, overlooking low-resource challenges and RH-ROP interplay. To address these gaps, we followed PRISMA guidelines,³ searching PubMed, Scopus, and Web of Science for peer-reviewed articles in English (2024–2025) using keywords like “neonatal retinal hemorrhage” and “ROP” combined with “diagnosis,” “screening,” or “artificial intelligence.” Studies were included if they addressed RH/ROP diagnosis or management in preterm infants with clear methodology, excluding case reports or non-neonatal studies. Recent 2024–2025 evidence highlights neonatal sepsis and oxygen therapy as key ROP risk factors,⁴ emphasizing the need for standardized screening and AI-driven solutions.⁵ By synthesizing these findings, this review aims to optimize early intervention, mitigate dataset biases in AI models, and enhance access to care in diverse settings, improving visual outcomes for preterm infants.

Current challenges in diagnosing neonatal retinal hemorrhage

Despite significant advancements in screening technologies, the diagnosis of neonatal RH continues to present considerable challenges. The primary diagnostic methods currently employed, including indirect ophthalmoscopy and wide-field digital retinal imaging systems, each have inherent limitations that impact the reliability and consistency of RH detection.⁶

Limitations of existing screening methods

Indirect ophthalmoscopy, though widely accessible and relatively straightforward, is heavily reliant on the clinician’s expertise and subjective interpretation. This dependence introduces a substantial potential for variability and diagnostic inaccuracies. In addition, the limited field of view inherent to this method restricts its capacity to detect peripheral RHs, often resulting in underdiagnosis.⁷

Wide-field digital retinal imaging systems, such as the RetCam III, have significantly enhanced the scope and quality of neonatal retinal screening. These systems provide comprehensive, real-time retinal images that can be archived for longitudinal follow-up and facilitate remote consultations. However, the high cost of these devices and the requirement for specialized training present substantial barriers to their widespread implementation, particularly in resource-limited settings.⁸

Variability in detection rates

The detection rate of RH is highly variable, influenced by both the timing of the initial screening and the diagnostic method employed. Early postnatal screenings tend to reveal higher incidences of RH, as hemorrhages present shortly after birth may be rapidly reabsorbed. For instance, recent studies report a 24.5% detection rate in newborns, with many hemorrhages resolving within a month.⁹ Similarly, 2025 observations show that 86% of RHs resolve within 2 weeks, highlighting the need for timely screening.¹⁰

The choice of screening method also contributes to detection rate variability. Recent studies demonstrate that digital imaging systems like RetCam significantly outperform indirect ophthalmoscopy in detecting neonatal RH in high-risk neonates, with higher sensitivity and broader field of view, detecting RH in approximately 20%–25% of preterm infants compared to lower rates with traditional methods.¹¹ These disparities underscore the necessity for standardized, comprehensive screening protocols to ensure consistent and accurate diagnosis across different clinical settings.

Need for standardized diagnostic criteria

The absence of universally accepted diagnostic criteria for neonatal RH complicates both clinical practice and research. Existing classification systems, such as those developed by von Barsewisch and Egge, primarily categorize hemorrhages by number and size but fail to adequately reflect the clinical severity and potential visual impact. Watts et al.¹² proposed a more nuanced classification system incorporating hemorrhage location and severity, but further validation and widespread adoption of this system are required.

The lack of standardized diagnostic criteria not only impedes clinical diagnosis and management but also hampers the comparability of research studies. Establishing a consensus on diagnostic standards is crucial to enhancing the consistency of RH diagnosis, improving clinical outcomes, and advancing research in this area.

Impact of delivery methods and neonatal conditions

Vaginal delivery versus cesarean section

The incidence of neonatal RH is significantly influenced by delivery method. Consistent evidence demonstrates higher RH rates following vaginal delivery compared to cesarean section.^13
–15 The use of forceps or vacuum extraction during vaginal delivery further elevates the risk of RH, with reported incidences markedly higher than in spontaneous vaginal deliveries.¹⁶ The mechanical forces and rapid pressure changes experienced by the infant during vaginal delivery contribute to the increased incidence of RH by elevating intracranial venous pressure, leading to retinal vessel rupture and hemorrhage.¹⁷ The compression of the fetal head during vaginal delivery causes a sharp increase in intracranial pressure and peripheral central retinal vein congestion, expansion, or rupture. Research indicates that RH is more prevalent in infants born through normal vaginal delivery compared to Cesarean section. The compression of the newborn’s head during vaginal delivery results in elevated intracranial venous blood pressure, impaired venous blood return, and subsequent rupture and bleeding of optic nerve endings. In addition, birth-related RH is frequently observed, especially in normal vaginal deliveries.¹⁸ The sudden rise in intracranial pressure during vaginal delivery, caused by the compression of the fetal head, leads to a sudden increase in intracranial pressure, affecting the retinal vasculature and resulting in retinal capillary hemorrhage.¹⁹

Furthermore, the rise in intracranial pressure can cause a reflux through glymphatic channels into the globe, leading to intraocular hemorrhage.²⁰ Studies have shown that vaginal delivery compression and the use of forceps may be associated with hemorrhage in preterm infants.²¹ The pattern of delivery and maternal conditions during pregnancy can also impact the prognosis of ROP.²² In cases of abusive head trauma, subdural hematoma in infants is more likely to be associated with retinal hemorrhage and cerebral edema.²³

In conclusion, evidence from various studies supports the notion that the mechanical forces and pressure changes experienced during vaginal delivery contribute to retinal hemorrhage in infants. The compression of the fetal head leads to increased intracranial pressure, venous congestion, and eventual rupture of retinal vessels, highlighting the importance of understanding and managing the risks associated with delivery methods to prevent such complications.

Influence of prematurity and associated conditions

Prematurity is a well-documented risk factor for RH. Preterm infants are particularly vulnerable due to their fragile retinal vasculature and the various complications associated with prematurity, such as respiratory distress syndrome and intraventricular hemorrhage.²⁴ The incidence of RH in preterm infants has been reported to be significantly higher than in full-term infants, with studies indicating that up to 31.8% of preterm infants exhibit RH within the first 24 h of life.²⁵ In addition, other neonatal conditions, such as neonatal asphyxia, hyperbilirubinemia, and neonatal infections, have been associated with an increased risk of RH. The risk of ROP in neonates is associated with various neonatal conditions such as sepsis, hyperbilirubinemia, and infections. Studies have shown that neonates with sepsis are at an increased risk of developing ROP²⁶ and that neonatal sepsis is one of the contributing factors to the incidence of ROP.²⁷ In addition, unstable clinical courses in neonates, including conditions like bronchopulmonary dysplasia, have been linked to ROP.²⁸ Furthermore, hyperbilirubinemia, a common condition in neonates, has been identified as a potential risk factor for ROP.²⁹ These findings emphasize the importance of monitoring and managing neonatal conditions to reduce the risk of ROP and its associated complications. The evidence highlights the critical impact of prematurity and associated neonatal conditions on the incidence of RH and ROP. The fragile retinal vasculature in preterm infants and the complications arising from prematurity significantly contribute to the higher rates of RH observed in these infants. In addition, conditions such as sepsis, hyperbilirubinemia, and bronchopulmonary dysplasia further exacerbate the risk, underlining the necessity for vigilant monitoring and comprehensive management strategies to mitigate these risks.

Correlation with retinopathy of prematurity

ROP is a significant concern in preterm infants and often coexists with RH. The underdeveloped retinal vasculature in these infants predisposes them to ROP, a vasoproliferative disease that complicates the clinical course.³⁰ Studies have demonstrated a correlation between the severity of RH and the development of ROP, indicating that severe RH significantly increases the risk of ROP in infants.³¹ This association underscores the necessity for vigilant ophthalmologic monitoring and timely interventions to mitigate adverse visual outcomes.²⁶ Recent research has also explored the potential impact of probiotics on the development of ROP in preterm infants, suggesting a beneficial relationship.³² Furthermore, advancements in technology, such as automated deep learning platforms for ROP screening using retinal images, have shown promising results, achieving diagnostic performance comparable to that of ophthalmologists.³³ In addition, studies investigating the impact of ocular pigmentation on retinal imaging in preterm infants highlight the importance of considering these factors during ROP examinations.³⁴

The association between neonatal RH and ROP is underpinned by shared pathophysiological mechanisms, particularly involving vascular endothelial growth factor (VEGF) dysregulation and perinatal oxidative stress pathways. In preterm infants, the immature retinal vasculature is highly susceptible to oxygen fluctuations, leading to a biphasic progression of ROP.³⁵ During phase I (hyperoxia), supplemental oxygen suppresses VEGF expression through mechanisms such as hypoxia-inducible factor-1α (HIF-1α) stabilization, resulting in vaso-obliteration and halted normal vascular development. This creates avascular zones in the peripheral retina, setting the stage for phase II (relative hypoxia), where VEGF is overexpressed, driving pathological neovascularization, increased vascular permeability, and potential fibrovascular proliferation.³⁶ RH exacerbates this process by causing acute vascular rupture and local ischemia, which further upregulates VEGF via inflammatory mediators and endothelial cell damage, amplifying the risk of severe ROP progression.^37,38

Perinatal oxidative stress pathways play a pivotal amplifying role in this interplay.³⁹ Preterm neonates, with underdeveloped antioxidant systems (e.g., reduced superoxide dismutase and glutathione peroxidase), are exposed to hyperoxia during neonatal care, generating excessive reactive oxygen species (ROS). These ROS induce endothelial cell apoptosis, disrupt tight junctions in retinal vessels, and activate proinflammatory cascades such as nuclear factor-kappa B (NF-κB), which in turn potentiate VEGF overexpression and vascular instability.⁴⁰ In the context of RH, hemorrhage-derived heme and iron further fuel oxidative stress through Fenton reactions, producing additional hydroxyl radicals that damage retinal tissues and promote a hypoxic microenvironment conducive to ROP. Recent studies highlight how targeting these pathways—such as with antioxidants, such as vitamin E or N-acetylcysteine—could mitigate VEGF-driven neovascularization and reduce RH-associated complications.⁴¹

Integrating these mechanisms, RH acts as an early sentinel event, indicating heightened vulnerability to oxidative stress and VEGF imbalance, which may accelerate ROP from mild to treatment-requiring stages. This underscores the potential for biomarkers such as serum VEGF levels or oxidative stress markers (e.g., malondialdehyde) to guide personalized screening and interventions in at-risk infants.

Given the correlation between neonatal RH and ROP, the presence of RH has significant implications for optimizing ROP screening strategies, particularly regarding the timing and frequency of initial ophthalmologic examinations in at-risk neonates. The detection of RH, especially within the first 24–48 h post-birth, serves as an early indicator of retinal vascular stress and heightened ROP risk, necessitating earlier initiation of screening in preterm infants with RH. A systematic review of global ROP guidelines⁴² highlights that screening criteria vary by country and neonatal care quality, with no universal standards possible due to differences in preterm infant characteristics; for instance, in high-income countries such as the United Kingdom,⁴³ screening is recommended for infants ⩽1500 g birth weight or ⩽30 weeks gestational age, typically starting at 4–6 weeks postnatal age, while in low- and middle-income countries like India or Kenyan,^44,45 wider criteria (e.g., ⩽2000 g or ⩽34 weeks) are used to account for higher ROP incidence in larger infants. However, infants with severe RH may benefit from earlier examinations (e.g., within 2–3 weeks) to detect preclinical ROP changes, as RH-induced ischemia can accelerate disease progression, a view supported by the review’s emphasis on adapting guidelines to local epidemiology, including risk factors such as RH. Furthermore, the severity and extent of RH may warrant more frequent follow-up examinations, such as biweekly instead of standard 1–2 week intervals, to monitor for rapid progression to treatment-requiring ROP stages; the review notes three strategies for initial timing—fixed postnatal intervals (e.g., 4 weeks in some guidelines), GA-dependent varying intervals (e.g., earlier for lower GA in UK or Australian guidelines), or mixed approaches—and suggests RH could shift toward more frequent GA-adjusted schedules.⁴⁶ Recent studies^47,48 suggest incorporating RH severity into risk stratification models to tailor screening schedules, potentially integrating digital retinal imaging to enhance early detection in high-risk neonates. These adjustments could improve timely interventions, reducing the risk of adverse visual outcomes in this vulnerable population.

Clinical significance and long-term outcomes

Visual development and potential complications

Neonatal RH and ROP present significant challenges due to their profound impact on visual development. Severe RH can result in structural damage to the retina, leading to long-term visual impairment. Preterm infants, with their underdeveloped retinal vasculature, are particularly susceptible to both RH and ROP. The spectrum of complications arising from these conditions can range from minor visual deficits to severe visual impairment, including blindness.⁴⁹

Long-term follow-up studies

Understanding the long-term impact of RH and ROP on visual outcomes necessitates extensive follow-up studies. Such studies have consistently demonstrated that infants who suffer from severe RH or advanced stages of ROP are at a heightened risk of enduring long-term visual and neurodevelopmental impairments. For example, research has shown that children who experienced severe RH during the neonatal period exhibit significantly higher rates of visual acuity deficits at age five compared to those without RH.⁵⁰ Likewise, infants with severe ROP frequently require surgical interventions such as laser therapy or vitrectomy. Although these procedures are effective in preventing blindness, they often result in suboptimal visual outcomes.⁵¹

Importance of early intervention

Early intervention is critical in mitigating the adverse outcomes associated with RH and ROP. Prompt diagnosis and timely treatment can markedly improve visual prognoses. The administration of anti-VEGF agents has proven effective in treating ROP by inhibiting abnormal retinal blood vessel growth. In addition, early screening and continuous monitoring of at-risk infants enable immediate intervention, which is crucial in preventing the progression of these conditions.⁸

Advancements in telemedicine and digital retinal imaging have further facilitated early detection and intervention, especially in remote or underserved areas. These technologies allow specialists to review retinal images and provide timely recommendations for treatment, ensuring that infants receive the necessary care without delay.⁵²

The clinical significance of RH and ROP extends well beyond the immediate neonatal period, with substantial long-term implications for visual development and overall quality of life. Comprehensive long-term follow-up and early intervention are vital in effectively managing these conditions and improving outcomes for affected infants.

Advances in diagnostic techniques

Role of digital retinal imaging

Digital retinal imaging has revolutionized the diagnosis of neonatal retinal conditions, including RH and ROP. These imaging systems, such as the RetCam, provide high-resolution, wide-field images of the retina, allowing for comprehensive visualization and accurate diagnosis. Digital imaging facilitates the documentation of retinal conditions, enabling longitudinal studies and telemedicine consultations, thus improving access to specialized ophthalmologic care in remote or resource-limited settings.³³

Comparison with traditional methods

Traditional methods of diagnosing RH and ROP, such as indirect ophthalmoscopy, while effective, have notable limitations. Indirect ophthalmoscopy requires a high level of expertise and is highly dependent on the examiner’s skill and experience, which can lead to variability in diagnosis. In addition, this method offers a limited field of view, often missing peripheral retinal lesions that are crucial for a comprehensive assessment.⁵³

In contrast, digital retinal imaging systems provide a much broader field of view and higher resolution images, significantly enhancing diagnostic accuracy. Studies have shown that digital imaging is superior in detecting retinal abnormalities, particularly in the peripheral regions, which are often missed during traditional examinations. Moreover, the ability to store and review images over time allows for better tracking of disease progression and treatment efficacy.

Future directions in technology

The future of diagnostic technology in neonatal retinal care is promising, with several advancements on the horizon. One significant development is the integration of artificial intelligence (AI) and machine learning algorithms with digital retinal imaging systems.⁵⁴ AI-powered platforms can analyze retinal images with high precision, offering diagnostic capabilities comparable to experienced ophthalmologists.⁵⁵ These systems can help in early detection and grading of ROP, potentially improving outcomes through timely intervention.^56,57

Furthermore, advancements in imaging technology, such as adaptive optics and optical coherence tomography (OCT), hold promise for even more detailed retinal examinations. These technologies provide high-resolution cross-sectional images of the retina, allowing for the detection of subtle structural changes that may precede clinical manifestations of RH and ROP.^58,59

In addition, portable and affordable retinal imaging devices are being developed to increase accessibility in low-resource settings. These innovations aim to bridge the gap in healthcare access, ensuring that even the most vulnerable populations receive timely and accurate diagnoses.⁶⁰

Beyond traditional AI, emerging technologies such as large language models (LLMs) and explainable AI (xAI) offer further potential for neonatal retinal care. LLMs can support clinical decision-making by analyzing literature or aiding risk stratification for RH and ROP, though their ophthalmic specificity requires validation.⁶¹ xAI enhances trust in AI-driven ROP screening by providing interpretable outputs, such as explaining vascular abnormalities in RH or ROP grading, critical for clinical adoption.⁶² However, challenges such as LLMs’ limited domain accuracy and xAI’s need for robust validation in diverse populations must be addressed to ensure effective integration.

Discussion

Neonatal RH and ROP present major ophthalmological challenges, with long-term effects on visual development, exacerbated by prematurity-related factors like vascular fragility, respiratory distress, and sepsis. Evidence shows a strong link between RH severity and ROP progression, necessitating strict monitoring for high-risk infants. Diagnosis remains problematic due to limitations in traditional methods, such as operator-dependent indirect ophthalmoscopy, and barriers to digital imaging adoption, including cost and training. The lack of standardized criteria further impedes research and clinical consistency. Technological advances, including digital imaging and AI, offer solutions by providing objective data and improving accuracy, efficiency, and accessibility through telemedicine. Modalities like OCT could enhance understanding of pathogenesis.

While AI holds transformative potential for ROP and RH management, its current limitations must be addressed to ensure effective clinical integration. AI-based systems for ROP and RH screening rely heavily on high-quality retinal images, which can be challenging to obtain in low-resource settings due to limited access to advanced imaging devices or skilled operators.⁶³ This is particularly critical for detecting subtle RH, where low-resolution images may miss small hemorrhages, increasing the risk of false negatives and delayed interventions. In addition, AI models often exhibit dataset bias, as most are trained on images from high-income countries, limiting generalizability to diverse populations, such as those in Latin America.⁶⁴ Interobserver variability among experts used for ground-truth labeling can further introduce inconsistencies, impacting model accuracy. Implementation challenges include high costs for system maintenance, staff training, and rigorous external validation, which may hinder adoption in resource-constrained environments.⁶⁵ To overcome these limitations, future efforts should focus on developing diverse datasets, improving image acquisition techniques like smartphone-based imaging, and establishing cost-effective validation protocols to enhance AI reliability and accessibility in clinical practice.

A multifaceted approach is essential, involving advanced screening, collaborative research for high-level evidence, and increased awareness for early detection. Integrating technology, standardized diagnostics, and vigilant care is key to better outcomes for affected infants.

Footnotes

Acknowledgements

We extend our gratitude to Sihang Liu from the University of Hong Kong for manuscript preparation assistance and English language editing.

Declarations

ORCID iD

Zhibin Xu

References

Srimanan

Kitsirilarp

Long-term ocular sequelae in preterm Thai infants: a comprehensive retrospective study. Medicine 2025; 104(6): e41485.

Liao

, et al. Risk factors for neonatal white-centered retinal hemorrhages: clinical implications for prevention. Int J General Med 2025; 18: 1747–1756.

Page

Mckenzie

Bossuyt

, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ Open Ophthalmol 2021; 372: n71.

Arya

Saini

Bhosale

, et al. Evaluation of the risk factors for retinopathy of prematurity in preterm infants admitted to the neonatal intensive care unit (NICU). Cureus 2025; 17(1): e77163.

Purohit

Usman

, et al. Physiological responses to retinopathy of prematurity screening: indirect ophthalmoscopy versus ultra-widefield retinal imaging. Pediatr Res. Epub ahead of print 13 February 2025. DOI: 10.1038/s41390-025-03906-4.

Choleva

Christodoulou

Herodotou

, et al. One-year review of retinopathy of prematurity in Cyprus: risk factors and screening effectiveness. J Pediatr Ophthalmol Strabismus 2025: 30: 1–7.

Simon-Szabó

Pál

Pándi

, et al. Retrospective analysis of the prevalence and risk factors of retinopathy of prematurity in a single-center cohort in romania and comparison with National and European Data. Medicina (Kaunas, Lithuania) 2025; 61(1): 149.

Bishnoi

Prasad

Upadhyay

, et al. A narrative review on managing retinopathy of prematurity: insights into pathogenesis, screening, and treatment strategies. Cureus 2024; 16(3): e56168.

Choi

Lee

, et al. Fundus photography-based distribution of retinal hemorrhages in newborns: implications for underlying mechanisms. J Personal Med 2025; 15(1): 38.

10.

Kabataş

Aydoğan

Bilgiç

, et al. Retinal hemorrhages and long-term ocular outcomes in neonatal hypoxic-ischemic encephalopathy. Medicina (Kaunas, Lithuania) 2025; 61(5): 906.

11.

Glaser

Härtel

Klingenberg

, et al. Neonatal sepsis episodes and retinopathy of prematurity in very preterm infants. JAMA Network Open 2024; 7(7): e2423933.

12.

Watts

Maguire

Kwok

, et al. Newborn retinal hemorrhages: a systematic review. J AAPOS 2013; 17(1): 70–78.

13.

Chen

Wang

, et al. Regional variations in retinopathy of prematurity incidence for preterm infants <32 weeks’ gestation in China. Public Health 2024; 226: 91–98.

14.

Icoz

Gündogan

Gundogan

K M

, et al. How is the posterior ocular segment affected in Cesarean section and normal vaginal delivery?. Ind J Ophthalmol 2024; 72(Suppl. 5): S893–S901.

15.

Choi

Lee

, et al. Fundus photography-based distribution of retinal hemorrhages in newborns: implications for underlying mechanisms. J Pers Med 2025; 15(1): 38.

16.

Pitti

Herling

, et al. Experimental assessment of traction force and associated fetal brain deformation in vacuum-assisted delivery. Ann Biomed Eng 2025; 53(4): 825–844.

17.

Lee-Confer

Wayman

Havens

KL.

High G-forces in unintentionally improper infant handling: implications for Shaken baby syndrome diagnosis. Forens Sci 2025; 5(2): 27.

18.

Kato

Nonaka

Akutsu

, et al. Correlations of intracranial pathology and cause of head injury with retinal hemorrhage in infants and toddlers: A multicenter, retrospective study by the J-HITs (Japanese Head injury of Infants and Toddlers study) group. PLoS One 2023; 18(3): e0283297.

19.

Al Zaabi

Naji

Tang

, et al. Utilising digital ocular imaging for paediatric retinal haemorrhages. Irish Med J 2025; 118: 3.

20.

Lin

Wang

Chen

, et al. Traumatic terson syndrome with a peculiar mass lesion and tractional retinal detachment: a case report. BMC Ophthalmol 2024; 24(1): 153.

21.

Hamou

Adhan

Lloyd

, et al. Retinal hemorrhage associated with neonatal idiopathic thrombocytopenia. Am J Ophthalmol Case Rep 2025; 38: 102310.

22.

Zhou

Cai

, et al. Clinical feature and predictive factor analysis for spontaneous regression of retinopathy of prematurity in a Chinese population. Int J Ophthalmol 2022; 15(12): 1978–1984.

23.

Nonaka

Asai

Abusive head trauma in infants and children in Japan. J Korean Neurosurg Soc 2022; 65(3): 380–384.

24.

Strauss

Januszkiewicz-Lewandowska

Sobaniec

, et al. SELENOP rs3877899 variant affects the risk of developing advanced stages of retinopathy of prematurity (ROP). Int J Mol Sci 2023; 24(8): 7570.

25.

Lin

W C

Jordan

B K

Scottoline

, et al. Oxygenation fluctuations associated with severe retinopathy of prematurity: insights from a multimodal deep learning approach. Ophthalmol Sci 2024; 4(2): 100417.

26.

Bonafiglia

Gusson

Longo

, et al. Early and late onset sepsis and retinopathy of prematurity in a cohort of preterm infants. Sci Rep 2022; 12(1): 11675.

27.

Boskabadi

Abrishami

Shoeibi

, et al. Incidence of retinopathy of prematurity in neonates with neonatal sepsis. Arch Iran Med 2023; 26(5): 248–51.

28.

Sherief

S T

Taye

Teshome

, et al. Retinopathy of prematurity among infants admitted to two neonatal intensive care units in Ethiopia. BMJ Open Ophthalmol 2023; 8(1): e001257.

29.

Fekri

Ojaghi

Momeni

, et al. Retinopathy of prematurity in Ardabil, North West of Iran: prevalence and risk factors. Eur J Translational Myol 2021; 31(4): 10063.

30.

Alghamdi

Alzahrani

Algarni

, et al. Retinopathy of prematurity in neonatal intensive care unit at King Fahd University Hospital, Eastern Province: prevalence, risk factors, and pattern of severity seven years’ experience. FResearch 2024; 13: 516.

31.

Shimoji

Suehiro

Matsuno

, et al. Abusive head trauma in Japan. Child’s nervous system: ChNS 2022; 38(12): 2387–2393.

32.

Bai

Dai

, et al. The relationship between probiotics and retinopathy of prematurity in preterm infants: A population-based retrospective study in China. Front Pediatr 2023; 11: 1055992.

33.

Wang

Zhang

, et al. Automated explainable multidimensional deep learning platform of retinal images for retinopathy of prematurity screening. JAMA Network Open 2021; 4(5): e218758.

34.

Seely

Mccall

Ying

, et al. Ocular pigmentation impact on retinal versus choroidal optical coherence tomography imaging in preterm infants. Transl Vision Sci Technol 2023; 12(7): 7.

35.

Fevereiro-Martins

Marques-Neves

Guimarães

, et al. Retinopathy of prematurity: a review of pathophysiology and signaling pathways. Survey Ophthalmol 2023; 68(2): 175–210.

36.

Kaur

Mikes

. Retinopathy of prematurity. In: StatPearls. Treasure Island (FL): StatPearls Publishing.

37.

Sjöbom

Hellqvist

Humayun

, et al. Circulating VEGF-A levels in relation to retinopathy of prematurity and treatment effects: a systematic review and meta-analysis. Ophthalmol Sci 2024; 4(6): 100548.

38.

Zhang

Buonfiglio

FIEß

, et al. Retinopathy of prematurity-targeting hypoxic and redox signaling pathways. Antioxidants (Basel, Switzerland) 2024; 13(2): 148.

39.

Hartnett

ME.

Pathophysiology of retinopathy of prematurity. Ann Rev Vision Sci 2023; 9: 39–70.

40.

Kimyon

Mete

Seyyar

SA.

Retinopathy of prematurity screening: can the examination time be reduced?

Clin Exper Optom 2025; 108(6): 729–732.

41.

Filippi

Gulden

Cammalleri

, et al. Retinopathy of prematurity in the era of precision neonatology: from risk stratification to targeted therapies. World J Pediatr 2025; 21(5): 430–435.

42.

Kościółek

Kisielewska

Ćwiklik-wierzbowska

, et al. Systematic review of the guidelines for retinopathy of prematurity. Eur J Ophthalmol 2022; 18: 11206721221126286.

43.

Aulakh

Houtman

A C

Rathod

, et al. The Royal College of Paediatrics and Child Health Retinopathy of Prematurity Screening Guidelines (2022): a series of treated infants falling outside the updated criteria. Eye 2024; 38(13): 2557–2560.

44.

Bowe

Nyamai

Ademola-Popoola

, et al. The current state of retinopathy of prematurity in India, Kenya, Mexico, Nigeria, Philippines, Romania, Thailand, and Venezuela. Eur J Ophthalmol 2019; 25(4): 49.

45.

Yusra

SM.

Knowledge, attitude and practices towards retinopathy of prematurity among health workers in the new-born units of Kilifi County Hospitals. University of Nairobi, 2023.

46.

Smith

A F

Sadeq

Kinzel

, et al. A systematic review of economic evaluations conducted for interventions to screen, treat, and manage retinopathy of prematurity (ROP) in the United States, United Kingdom, and Canada. Ophthalm Epidemiol 2023; 30(2): 113–120.

47.

Chen

Zhao

, et al. Multi-risk factors joint prediction model for risk prediction of retinopathy of prematurity. EPMA J 2024; 15(2): 261–274.

48.

Timkovicˇ

Nowaková

Kubícˇek

, et al. Retinal image dataset of infants and retinopathy of prematurity. Sci Data, 2024; 11(1): 814.

49.

Robitaille

JM.

Long-term visual outcomes in prematurely born children. J Binocul Vis Ocul Motil 2024; 74(1): 1–8.

50.

Koc

Bas

AY.

Retinopathy of prematurity (ROP): from the perspective of the neonatologist. Global Pediatr 2024; 8: 100159.

51.

Mintz-hittner

Kennedy

K A

Chuang

AZ.

Efficacy of intravitreal bevacizumab for stage 3+ retinopathy of prematurity. N Engl J Med 2011; 364(7): 603–615.

52.

Padhi

Bhunia

Das

, et al. Outcome of real-time telescreening for retinopathy of prematurity using videoconferencing in a community setting in Eastern India. Indian J Ophthalmol 2024; 72(5): 697–703.

53.

Chew

Gadiraju

Saini

, et al. Pediatric eye screening: current standards and gaps in care. Ophthalmic Surg Lasers Imaging Retina 2025; 56(4): 232–239.

54.

Jafarizadeh

Maleki

S F

Pouya

, et al. Current and future roles of artificial intelligence in retinopathy of prematurity. Artif Intell Rev 2025; 58(6): 1–55.

55.

Ortiz

Patiño

Torres

, et al. AI-enabled screening for retinopathy of prematurity in low-resource settings. JAMA Netw Open 2025; 8(4): e257831-e.

56.

Liu

Xie

Zhao

, et al. Automated detection of nine infantile fundus diseases and conditions in retinal images using a deep learning system. EPMA J 2024; 15(1): 39–51.

57.

Durmaz Engin

Ozturk

Ozkan

, et al. Prediction of retinopathy of prematurity development and treatment need with machine learning models. BMC Ophthalmol 2025; 25(1): 1–12.

58.

Chen

Tsui

, et al. Insights into the developing fovea revealed by imaging. Progr Retinal Eye Res 2022; 90: 101067.

59.

Mujat

Ferguson

Hammer

, et al. High-resolution retinal imaging: technology overview and applications. Photonics 2024; 11(6): 522.

60.

Ahmed

Shah

S V

Sedhom

, et al. Pediatric telemedicine-based retinal screening in Ghana. Medknow 2025; 8(1): 5–8.

61.

Oniani

Visweswaran

, et al. Enhancing large language models for clinical decision support by incorporating clinical practice guidelines. In: Proceedings of the 2024 IEEE 12th International Conference on Healthcare Informatics (ICHI), 2024. New York: IEEE.

62.

Asif

Ur Rehman

Rashid

, et al. An insight on the timely diagnosis of diabetic retinopathy using traditional and AI-driven approaches. IEEE Access 2025; 13.

63.

Tsai

Yip

Song

, et al. Implementation of artificial intelligence in retinopathy of prematurity care: challenges and opportunities. Int Ophthalmol Clin 2024; 64(4): 9–14.

64.

Coyner

Murickan

, et al. Multinational external validation of autonomous retinopathy of prematurity screening. JAMA Ophthalmol 2024; 142(4): 327–335.

65.

Rashidian

Karami

Salehi

SA.

A review on retinopathy of prematurity. Med Hypothesis Discov Innov Ophthalmol 2025; 13(4): 201.