MediAssist-AI: Cardiovascular Risk Assessment
AI-powered CVD risk prediction with clinical interpretability
The Problem
Traditional CVD risk assessment tools lack transparency in their predictions. Clinicians need systems that not only predict risk but explain why a patient is at risk, enabling better patient communication and clinical decision-making.
What I Built
Built a hybrid AI system combining a weighted ensemble of four boosting algorithms (Random Forest, LightGBM, XGBoost, CatBoost) with 8 clinical rules based on ACC/AHA and WHO guidelines. The system achieves 96.54% recall to minimize missed CVD cases.
My Role
Led the development of the ML pipeline, ensemble model architecture, and clinical rule integration. Built the Streamlit dashboard and Dr. CardioExpert chatbot.
How It Works
Weighted voting ensemble combining 4 ML algorithms
22 engineered features from 11 patient health metrics
8 clinical rules based on ACC/AHA and WHO guidelines
Streamlit dashboard with dark/light mode support
Dr. CardioExpert AI chatbot for patient education
Real-time predictions with visual risk indicators
Key Takeaways
High recall is critical in medical AI to minimize missed diagnoses
Clinical interpretability builds trust with healthcare professionals
Ensemble methods improve robustness over single models
Rule-based reasoning complements ML for explainability