We are building a framework that harmonizes clinical records, imaging, genomics, physiological signals and lifestyle data around a cardiovascular knowledge graph, and applies graph neural networks, causal models and language models to predict risk, disease progression and response to treatment for the individual patient.
Data in
- Clinical records: history, notes, labs, medications
- Echocardiography, MRI, CT, ultrasound and angiography
- ECG, PPG, blood pressure and heart-rate variability
- Genomics, proteomics and lifestyle factors
What it is meant to deliver
- Cardiovascular risk prediction and early warning
- Disease subtyping and progression modeling
- Personalized treatment recommendations
- Population-level identification of high-risk groups
Overview
The full picture.
From one patient's heart data to decisions for that patient, and to prevention for whole populations. Planned framework; no results are reported here.
01Cardiovascular data
- Clinical data (EHR)history, notes, labs, medications, outcomes
- Genomics & multi-omicstranscriptomics, proteomics, metabolomics, pathways
- Imagingechocardiography, MRI, CT, ultrasound, angiography, radiomics
- Physiological signalsECG, PPG, blood pressure, HRV, remote monitoring
- Lifestyle & environmentdiet, activity, stress, sleep, air pollution, socioeconomic factors
02Harmonize & connect
- Standardization & de-identification
- Cross-modal fusion
- Real-time & longitudinal data
- Cardiovascular knowledge graphphenotypes, biomarkers, mechanisms, guidelines, patient similarity
03AI-driven CVD intelligence
CVD intelligenceReasoning · prediction · personalization
Models
- Graph neural networks
- Neuro-symbolic & causal AI
- Language & multimodal models
- Patient & disease digital twin
Predictive modeling
- Early detection
- Disease progression
- Treatment response
- What-if analysis
04Care & prevention
- Clinical decision supportrisk prediction, early warning, subtyping, explainable insights
- Personalized managementtargeted therapy, lifestyle advice, remote monitoring and alerts
- Population health & precision preventionrisk stratification at scale, high-risk groups, equitable care
✓Trustworthy, responsible & secure AI
- Privacy & data security
- Fairness & bias mitigation
- Explainability
- Human in the loop
- Robustness & reliability
- Regulatory compliance & ethics
- Earlier detection
- Personalized care
- Improved outcomes
- Scalable & equitable solutions
- Healthier communities
Illustrations generated with ChatGPT for the AI4Health Lab. They depict research themes, not study data.