Assistant Professor Baylor University Waco, Texas, United States
Introduction: This project explores creating and deploying a dual biosensor system to gather a patient’s cardiac signals and compute the Stroke Volume Allocation (SVA) model to compare it and the physiological signals to an existing clinical database. The system captures and analyzes essential cardiovascular metrics like BPM (beats per minute), BP (blood pressure), and sound waves gathered near the heart, sending this data to a local computer application. The aim is to train an AI model on sets of existing clinical data where the integration of cloud-based data processing ensures that large datasets can be managed and analyzed efficiently, enhancing the predictive capabilities of the AI model. The model will compare these data sets to the collected and calculated data to provide more accurate cardiovascular health assessments and a comprehensive evaluation of a patient’s heart valve.
Materials and
Methods: Two types of biosensors were developed when designing the biosensors: a wrist-worn sensor and an acoustic sensor placed near the heart. The wrist-worn sensor is modeled after a smartwatch to collect a user’s BPM. The acoustic sensor mimics the design of a stethoscope and collects a user’s sound waves gathered from the heart. It does so effectively using a bi-directional microphone where one mic is facing into the body, and the other is seated behind the other and facing outwards. This system is used to enhance noise cancellation and assist with isolating the sound waves from the heart. Both biosensors continuously collect physiological data for 15 minutes and integrate it with a single BP measurement taken by a Beurer BM55. The collected data undergoes preprocessing in a local application before being uploaded to a cloud server. In the cloud, additional cleaning and processing is done to improve the data’s quality and prepare it for detailed analysis. This process includes calculating the SVA model, which provides metrics such as arterial stiffness, elasticity, vascular age, and cardiovascular disease risk. In parallel with this analysis, the physiological data and SVA model are compared against clinical datasets to train a machine learning model. Various algorithms optimize this model to generate accurate predictions for cardiovascular health outcomes. The results from these models are then sent to an inference engine, which synthesizes the data to offer actionable health insights within the computer application for the user.
Results, Conclusions, and Discussions: The expected results for this project include a well-defined biosensor system with data collection similar to those that meet medical grade and a practical and straightforward AI model that can be used to evaluate a patient’s cardiovascular health and their heart valve’s efficiency. With those results in mind, it is possible for this technology to be integrated into existing biosensors commonly used by the public, such as smartwatches. Additionally, this project aims to assist in current medical care where the adoption of the digital stethoscope is currently taking place. Finally, the prospect of quick and efficient health assessment can be used in remote areas to evaluate those who may not be able to reach a hospital. Overall, this research has tremendous promise and could lead to earlier detection of heart-related issues and more personalized treatment plans, potentially reducing the incidence of severe cardiac events and improving patient outcomes.
.jpg "Barrett L. Burgess, BS (he/him/his) photo")
.jpg "Taeil Kim, PhD photo")