Complementary Roles of Image Classification and Vessel Segmentation in AI-Based Screening for Retinopathy of Prematurity Plus Disease in a Kenyan Preterm Cohort

2026-07-08 · Source: cs.CV updates on arXiv.org · Field: Health & Wellbeing — Medical Devices & Health Technology, Clinical Care & Medical Practice, Health & Medical Research · Depth: Expert, quick

Summary

A study on AI-based screening for Retinopathy of Prematurity (ROP) Plus disease in a Kenyan preterm cohort evaluated the complementary roles of image classification and vessel segmentation. Analyzing 1,635 fundus images from 237 eyes of 121 Kenyan preterm infants, researchers tested eleven configurations, including image classifiers and multi-task segmentation-classification pipelines. Vessel segmentation proved feasible, achieving a Dice score of 0.533 and specificity of 0.979. While RGB classifiers showed high sensitivity but high over-referral rates, segmentation-coupled models improved specificity. The most balanced performance came from a probability ensemble combining both approaches, yielding a sensitivity of 0.692, specificity of 0.914, and balanced accuracy of 0.803, surpassing standalone vision classifiers. This highlights the benefit of integrated workflows for ROP Plus detection in resource-limited settings.

Key takeaway

For AI scientists developing diagnostic tools for Retinopathy of Prematurity (ROP) Plus disease, particularly in resource-limited regions, your strategy should integrate both image classification and vessel segmentation. Relying solely on classifiers risks high over-referral rates, while segmentation improves specificity. You should prioritize combined workflows, such as probability ensembles, to achieve balanced performance, and ensure your systems undergo prospective multi-site validation for real-world applicability.

Key insights

Image classification and vessel segmentation offer complementary strengths for AI-based ROP Plus disease detection.

Principles

• Classifiers excel at sensitive case-finding.
• Segmentation enhances specificity and reduces over-referral.
• Combined AI workflows improve diagnostic accuracy.

Method

Evaluation involved patient-grouped nested cross-validation across 11 configurations, including image classifiers, multi-instance learning, multi-task segmentation-classification, and segment-then-classify pipelines.

In practice

• Implement OR-based screening for high sensitivity.
• Use AND-based confirmation for high specificity.
• Employ probability ensembles for balanced performance.

Topics

• Retinopathy of Prematurity
• Plus Disease
• Image Classification
• Vessel Segmentation
• AI Medical Diagnostics
• Fundus Imaging
• Kenyan Healthcare

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• Complementary Roles of Image Classification and Vessel Segmentation in AI-Based Screening for Retinopathy of Prematurity Plus Disease in a Kenyan Preterm Cohort — Combining image classification and vessel segmentation improves AI-based ROP Plus detection by balancing sensitivity and specificity.
• A Dual Edge Spatial Jacobian Image Graph for Interpretable Diabetic Retinopathy Grading — Interpretable DR grading benefits from graph representations linking spatial lesion-vessel geometry and embedding-biomarker sensitivity.
• Cross-Source Supervision for Bone Infection Segmentation in Dual-Modality PET-CT — Decoupled dual-source learning with PET-CT fusion improves bone infection segmentation by accommodating diverse clinical annotation philosophies.
• Could AI Imaging Spot Pregnancy Health Risks Earlier? — AI can significantly enhance early detection of high-risk pregnancy conditions like PAS and CHD, improving care pathways.
• Taxlifier: Leveraging Disease Taxonomy for Enhanced Multi-Label Classification in Chest Radiography — Leveraging disease taxonomy in hierarchical multi-label classification significantly improves chest X-ray diagnosis accuracy and interpretability.
• Trusted Multi-View Deep Learning Classification of Fetal Congenital Heart Disease with Feature-level and Decision-level Fusion — The framework integrates multi-view echocardiographic data with uncertainty handling for robust fetal CHD classification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.