About
Overview: Children with congenital heart disease face significant risks during hospital stays, with unplanned ICU readmissions occurring in 4-20% of pediatric cardiac patients, rates are substantially higher than the general pediatric population. These deterioration events lead to increased length of stay, higher costs, and potential psychological trauma for both patients and families. Current prediction models have limited accuracy and provide little actionable guidance to clinicians. Our research team is developing and evaluating an innovative AI system that combines advanced prediction capabilities with explainable clinical reasoning. Unlike existing warning systems that simply provide risk scores, our system generates contextualized explanations for each alert, helping clinicians understand why a patient might be at risk and what similar cases have shown in the past.
Aims: We will be enhancing the prediction model through multimodal data integration, incorporating ECGs, imaging data, and real-time physiological monitoring alongside traditional clinical data. Using reinforcement learning from human feedback, the system will evolve based on real-world usage patterns and clinician input. The system will undergo rigorous testing through silent trials followed by clinical deployment. We will measure predictive performance, clinician adoption, user satisfaction, and ultimately impact on patient outcomes.
Impact: This work aims to improve outcomes for children with heart disease while advancing the field’s understanding of how generative AI can be safely and effectively integrated into clinical workflows. The research will inform future guidelines for AI deployment in pediatric care and contribute to the development of more transparent, trustworthy clinical decision support systems.
Accepting Students
This multidisciplinary project offers multiple opportunities for students (both MSc and PhD) in computer science, clinical informatics, bioethics, and implementation science. Students will work with cutting-edge AI technologies while addressing real clinical challenges that directly impact patient care. The research environment emphasizes responsible AI development with robust safety protocols, bias detection strategies, and equity considerations. Students will gain experience in multimodal machine learning, natural language processing, clinical data science, and human-computer interaction in healthcare settings. Collaboration opportunities exist across multiple institutions and disciplines, with mentorship from experts in pediatric cardiology, AI research, clinical ethics, and health services research. The project’s staged implementation approach ensures students can contribute meaningfully while learning about the complexities of translating AI research into clinical practice.
