Can we find hidden heart risk in people who seem healthy? Cardiovascular disease remains the leading cause of death globally. Yet standard screenings often miss people who appear healthy , young, normal weight, no symptoms. Could some of these low-risk individuals actually be at hidden risk? My project focuses precisely on this subgroup. Using real biomedical data, I trained a machine learning model only on healthy-looking individuals to detect hidden cardiovascular risk. The results were striking: 34% of the low-risk group showed signs of disease, despite standard tests marking them healthy. I went further: using SHAP explainability, I revealed which subtle factors drive this hidden risk. This makesthe model not just predictive, but interpretable and actionable for doctors.
What I Learned
Working with real biomedical data.
Training ML models for subtle, low-risk patterns.
Making AI predictions interpretable with SHAP.
Communicating complex medical insights clearly.
How I Built It
Selected “healthy-looking” individuals from biomedical datasets.
Trained a Gradient Boosted Decision Tree to predict hidden risk.
Challenges
Small, imbalanced subgroup → solved with sampling and class weighting.
Feature correlations → iterative selection.
Explaining SHAP outputs in simple terms.
Impact The project shows that healthy-looking individuals can still be at risk. By revealing which factors matter, it supports early preventive action and empowers people to protect their heart health.
Log in or sign up for Devpost to join the conversation.