Unified Diagnostic Intelligence: RESTful Integration of Validated Clinical Data and Machine Learning for Heart Health Prediction

Abstract

Neha Bansal and Bhawna Singla

Heart disease remains one of the leading causes of death globally, with early prediction being crucial for timely intervention and improved patient outcomes. However, existing predictive systems are often constrained by fragmented, non-standardized clinical data across diagnostic centers. This paper presents a RESTful framework for heart disease prediction that leverages validated clinical data and state-of-the-art machine learning (ML) models. The system integrates data collection via a user-friendly website, validation using Pydantic models, and a FastAPI-based RESTful API for scalable, asynchronous data ingestion. Data undergoes rigorous validation at both frontend and backend levels before being preprocessed and transformed for machine learning.

The ML pipeline includes imputation, scaling, and encoding of numeric and categorical features, followed by model selection using stratified 5-fold cross-validation on RandomForest and XGBoost classifiers. The best-performing model is trained on the entire dataset, and both the preprocessing pipeline and model are saved using joblib for reproducibility and inference. While the framework currently operates on synthetically generated data, it demonstrates the enormous potential of predictive modeling in healthcare. With institutional cooperation in sharing real, anonymized clinical data, this approach can significantly enhance the generalizability and accuracy of heart disease prediction, thereby transforming diagnostic decision support and ultimately benefiting public health outcomes.

PDF