Inspiration
Heart disease is the leading global cause of death, yet early detection remains challenging. Our motivation was to empower patients and doctors with quick, interpretable AI-driven diagnosis to save lives.
What it does Our project leverages Google’s advanced Gemini AI API integrated with a machine learning model to predict heart disease risk rapidly and accurately. By combining a Random Forest classifier trained on clinical data with Gemini’s natural language processing, our web app not only provides reliable predictions but also delivers easy-to-understand explanations of the results. This empowers patients and healthcare providers with actionable insights while enhancing trust through transparent AI.
What it does
HeartDiseaseDetector predicts heart disease risk from clinical data using a machine learning model enhanced by Google’s Gemini API. It explains results in natural language, making complex AI predictions understandable and actionable.
How we built it
We trained a Random Forest classifier on a clean heart disease dataset, preprocessing numerical and categorical features, scaling numeric attributes, and then integrated the model into a Flask web app. Gemini API generates clear, empathetic explanations for predictions.
Challenges we ran into
Handling categorical vs. numeric features during preprocessing was tricky. Integrating Gemini API required careful prompt design to generate helpful and precise explanations while managing API limits and latency for real-time user interaction.
Accomplishments that we're proud of
Achieving over 87% prediction accuracy and seamless natural language explanations through Gemini AI. The model’s interpretability and real-time deployment in a web app demonstrate a powerful fusion of ML and generative AI for healthcare.
What we learned
The importance of explainable AI in clinical settings, effective data preprocessing practices, and integrating large generative models like Gemini into practical applications. Also, how critical user-friendly explanations are for patient trust.
What's next for HeartDiseaseDetector
Expand to multi-disease prediction using additional health datasets, improve the explanation module with patient-tailored advice, and deploy on scalable cloud platforms with enhanced privacy safeguards for real-world clinical use.
Log in or sign up for Devpost to join the conversation.