Design of a Tiny ML-Based Predictive Heart Disease Screening Device

Authors- Acquah Andrews, Quaye Nii Attoh Gabriel

Abstract- -Cardiovascular disorders have become a global issue. There is a prevalence of heart failure and its impact on public health, particularly in regions with limited healthcare resources such as Africa. To combat this challenge, governments and healthcare institutions have initiated awareness campaigns and infrastructure improvements. Technological advancements, particularly in machine learning, offer promising solutions for early detection and prognosis. This study explores the application of Tiny Machine Learning (TinyML) in the context of heart disease, leveraging its potential for quantized models on resource-constrained devices. The research examines traditional diagnostic methodologies and highlights the application of TinyML in predicting heart failure. The research contributes a unique perspective by deploying a Shallow Neural Network (SNN) model on an Arduino BLE 33 Sense for heart failure prediction, focusing on eight features. The resulting classification report demonstrates the model’s accuracy of 82.61% and ROC_AUC of 92.15% for both absence and presence of heart disease. This paper serves as a foundation for future enhancements and applications in predictive healthcare technologies.

DOI: 10.61463/ijset.vol.11.issue6.103