Artificial intelligence for patent ductus arteriosus-a systematic review

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence

Background:

• Optimal management of patent ductus arteriosus (PDA) is complex and controversial.
• High-dimensional data and the need for individualized assessment highlight PDA as a candidate for AI-driven approaches.

Purpose of the Study:

• To systematically review Artificial Intelligence (AI) research in patent ductus arteriosus (PDA).
• To identify strengths, limitations, and future directions of AI applications in PDA management.

Main Methods:

• Systematic review following PRISMA 2020 guidelines, searching databases from January 1, 2010, to May 31, 2025.
• Extraction of data on study design, population, AI methods, performance, validation, and explainability.
• Assessment of risk of bias and reporting quality; narrative synthesis due to heterogeneity.

Main Results:

• Eleven studies were included, focusing on diagnosis, treatment response, risk factors, complications, and subphenotypes.
• Most studies were retrospective, single-center, with small sample sizes and varied definitions.
• AI models showed potential (AUCs 0.74-0.93 for diagnosis) but suffered from high risk of bias, overfitting, and inadequate external validation.

Conclusions:

• AI demonstrates feasibility for PDA risk stratification, diagnosis, and outcome prediction.
• Current AI applications are preliminary and not clinically ready; rigorous validation and bias-aware methods are needed.
• Future research must prioritize clinical utility, fairness, explainability, and reproducibility for successful translation.