[Paper Review] Current and future roles of artificial intelligence in retinopathy of prematurity

Ali Jafarizadeh, Shadi Farabi Maleki+10 more|arXiv (Cornell University)|Feb 15, 2024

Retinopathy of Prematurity Studies5 citations

TL;DR

This review surveys the use of retinal imaging and AI for detecting and managing retinopathy of prematurity (ROP), highlighting the shift from traditional ML to deep learning and outlining current progress, challenges, and future directions.

ABSTRACT

Retinopathy of prematurity (ROP) is a severe condition affecting premature infants, leading to abnormal retinal blood vessel growth, retinal detachment, and potential blindness. While semi-automated systems have been used in the past to diagnose ROP-related plus disease by quantifying retinal vessel features, traditional machine learning (ML) models face challenges like accuracy and overfitting. Recent advancements in deep learning (DL), especially convolutional neural networks (CNNs), have significantly improved ROP detection and classification. The i-ROP deep learning (i-ROP-DL) system also shows promise in detecting plus disease, offering reliable ROP diagnosis potential. This research comprehensively examines the contemporary progress and challenges associated with using retinal imaging and artificial intelligence (AI) to detect ROP, offering valuable insights that can guide further investigation in this domain. Based on 89 original studies in this field (out of 1487 studies that were comprehensively reviewed), we concluded that traditional methods for ROP diagnosis suffer from subjectivity and manual analysis, leading to inconsistent clinical decisions. AI holds great promise for improving ROP management. This review explores AI's potential in ROP detection, classification, diagnosis, and prognosis.

Motivation & Objective

• Assess limitations of traditional ML methods in ROP diagnosis due to subjectivity and overfitting.
• Summarize advancements in deep learning for ROP detection and plus disease identification.
• Evaluate AI-based systems (e.g., i-ROP-DL) for reliable ROP diagnosis.
• Identify challenges and gaps hindering translation of AI from research to clinical practice.
• Provide guidance for future research directions in AI-enabled ROP management.

Proposed method

• Systematic review of 89 original studies on AI and ROP from a larger set of 1487 studies.
• Synthesis of methods from traditional ML to CNN-based approaches for retinal image analysis.
• Discussion of performance, reliability, subjectivity, and clinical integration issues.
• Comparison of AI-driven detection, classification, diagnosis, and prognosis tasks.
• Critical appraisal of datasets, validation, and generalizability considerations.

Experimental results

Research questions

• RQ1 What are the current capabilities and limitations of AI methods for detecting and classifying ROP from retinal images?
• RQ2 How has deep learning, particularly CNNs, improved ROP diagnosis and plus disease detection compared to traditional ML?
• RQ3 What are the key challenges to clinical adoption of AI in ROP management (e.g., subjectivity, overfitting, validation)?
• RQ4 What future directions and research gaps exist to advance AI-assisted ROP care?
• RQ5 How does AI influence prognosis and decision-making in ROP management?

Key findings

• Traditional ROP diagnostic methods suffer from subjectivity and manual analysis leading to inconsistent decisions.
• DL approaches, including CNNs, have significantly improved detection and classification of ROP and plus disease.
• The i-ROP-DL system shows promise for reliable automatic ROP diagnosis.
• AI has strong potential to improve ROP management across detection, classification, diagnosis, and prognosis.
• There are notable challenges and gaps in data, validation, generalizability, and clinical translation that need addressing.
• The review is based on 89 original studies identified within a broader literature body.