Advancing Chronic Respiratory Disease Care with Real-Time Vital Sign Prediction

Abstract

Cardiovascular and chronic respiratory diseases, being pervasive in nature, pose formidable challenges to the overall well-being of the global populace. With an alarming annual mortality rate of approximately 19 million individuals across the globe, these diseases have emerged as significant public health concerns warranting immediate attention and comprehensive understanding. The mitigation of this elevated mortality rate can be achieved through the application of cutting-edge technological innovations within the realm of medical science, which possess the capacity to enable the perpetual surveillance of various physiological indicators, including but not limited to blood pressure, cholesterol levels, and blood glucose concentrations. The forward-thinking implications of these pivotal physiological or vital sign parameters not only facilitate prompt intervention from medical professionals and carers, but also empower patients to effectively navigate their health status through the receipt of pertinent periodic notifications and guidance from healthcare practitioners. In this research endeavour, we present a novel framework that leverages the power of machine learning algorithms to forecast and categorise forthcoming values of pertinent physiological indicators in the context of cardiovascular and chronic respiratory ailments. Drawing upon prognostications of prospective values, the envisaged framework possesses the capacity to effectively categorise the health condition of individuals, thereby alerting both caretakers and medical professionals. In the present study, a machine-learning-driven prediction and classification framework has been employed, wherein a genuine dataset comprising vital signs has been utilised. In order to anticipate the forthcoming 1-3 minutes of vital sign values, a series of regression techniques, namely linear regression and polynomial regression of degrees 2, 3, and 4, have been subjected to rigorous examination and evaluation. In the realm of caregiving, a concise 60-second prognostication is employed to enable the expeditious provision of emergency medical aid. Additionally, a more comprehensive 3-minute prognostication of vital signs is utilised for the same purpose. The patient's overall health is evaluated based on the anticipated vital signs values through the utilisation of three machine learning classifiers, namely Support Vector Machine (SVM), Decision Tree and Random Forest. The findings of our study indicate that the implementation of a Decision Tree algorithm exhibits a high level of accuracy in accurately categorising a patient's health status by leveraging anomalous values of vital signs. This approach demonstrates its potential in facilitating prompt and effective medical interventions, thereby enhancing the overall quality of care provided to patients.