AI in Pediatric CHD

Summary

This project explores the transformative role of artificial intelligence, Federated Learning, and Explainable AI in diagnosing and managing congenital heart diseases (CHD) in pediatric patients. It focuses on conditions such as Hypoplastic Left Heart Syndrome (HLHS), Pulmonary Hypertension (PAH), Aortic Stenosis, Patent Ductus Arteriosus (PDA), and Ventricular Septal Defect (VSD).
Key applications in clinical use cases include:
Echocardiography Data Anonymization: Ensuring patient privacy by anonymizing echocardiographic data, which facilitates secure and compliant use of sensitive information.
Disease Classification: Utilizing AI to quickly and accurately classify various types of CHDs, thus improving diagnostic efficiency and aiding in timely treatment.
Segmentation of Cardiac Structures: Employing AI for precise segmentation of cardiac structures in echocardiograms, which enhances visualization and supports better treatment planning.
Quantitative Assessment of Cardiac Function: Applying AI to perform detailed quantitative analyses of cardiac function, enabling more informed and effective clinical decisions.
Large Language Models: Creating language models to interpret medical guidelines and patient health records, improving the accessibility and understanding of complex medical information.
By leveraging these advanced AI technologies, especially in echocardiography, healthcare professionals can achieve more accurate diagnoses, develop personalized treatment plans, and enhance outcomes for children with CHD.

Publications:

• T. Uden, MY Jabarulla, S. Oeltze-Jafra, P. Beerbaum, "Fine-Tuning Language Models with Guideline Knowledge to Optimize Small, Local Models for Pediatric Cardiology" Accepted for Prentation in The Association for European Paediatric and Congenital Cardiology (AEPC) 2025, Hamburg, Germany.
• Sarah Elizabeth Long, Theodor Uden, Mohamed Yaseen Jabarulla, Steffen Oeltze-Jafra, Philipp Beerbaum"Automated Selection and Annotation of Unstructured Pediatric Echocardiography Reports UsingLlama-3.1-8B-Instruct, a Locally Run Large Language Model"Accepted for Prentation in The Association for European Paediatric and Congenital Cardiology (AEPC) 2025, Hamburg, Germany.
• H. Rosmus, T. Uden, MY Jabarulla, K. Mirmukhamedov, M. Fischer, C. M. Happel, D. Hohmann, S. Lohrmann, C. Junge, S. Long, J. Ulmer, S. Oeltze-Jafra, P. Beerbaum,"AI-based right ventricular segmentation in children with pulmonary hypertension ", Accepted for Oral Prentation in The Association for European Paediatric and Congenital Cardiology (AEPC) 2025, Hamburg, Germany.
• T Uden, MY Jabarulla, T Jack, M Avsar, H Bertram, C Happel, D Hohmann, A Horke, C Junge, S Long, N Rudobashta, K Seidemann, S Oeltze-Jafra, P Beerbaum 'Large Language Models Provide Impressive Answers to Complex Questions from Parents in Pediatric Cardiology and Pediatric Cardiac Surgery, But Detecting Errors is Challenging' The Thoracic and Cardiovascular Surgeon, 2025/1, 73(S 02) pp: DGPK-KV33. DOI: 10.1055/s-0045-1804267
• T Uden, Y M Jabarulla, T Jack, M Avsar, H Bertram, C M Happel, D Hohmann, A Horke, C Junge, S Long, N Rudobashta, K Seidemann, S Oeltze-Jafra, P Beerbaum, "A guideline-informed language model for paediatric cardiology demonstrates high performance in answering complex medical questions", European Heart Journal, 45, Issue Supplement_1, 2024 (Impact Factor: 37.9) DOI: ehae666.3491
• Mohamed Yaseen Jabarulla, Steffen Oeltze-Jafra, Philipp Beerbaum and Theodor Uden "MedDoc-Bot: A Chat Tool for Comparative Analysis of Large Language Models in the Context of the Pediatric Hypertension Guideline" Presented at the 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Orlando, Florida, USA, July 15-19, 2024. DOI: doi.org/10.1109/EMBC53108.2024.10781509.
• MY Jabarulla, T Uden, P Beerbaum, S Oeltze-Jafra "Artificial intelligence in pediatric echocardiography-Automated view classification and image anonymization in rare cardiac malformations on the example of borderline HLHS" European Heart Journal, 44, Issue Supplement_2, 2023 (Impact Factor: 37.9) DOI: doi.org/10.1093/eurheartj/ehad655.061

ArXiv Articles:

• MY Jabarulla, T Uden,T Jack, P Beerbaum, S Oeltze-Jafra "Artificial Intelligence in Pediatric Echocardiography: Exploring Challenges, Opportunities, and Clinical Applications with Explainable AI and Federated Learning" ArXiv. Link

Under Review

• MY Jabarulla, T Uden, P Beerbaum, S Oeltze-Jafra 'Efficient Fine-Tuning and Evaluation of Large Language Models for Interpreting Pediatric Hypertension Guidelines' submitted to EMBC 2025.

Softwares and Codes
MedDoc-Bot: Github Repository
Fine-Tunning LLM: Github Repository