Induced bioresistance via BNP detection for machine learning-based risk assessment

Abstract

Machine Learning (ML) is a powerful tool for big data analysis that shows substantial potential in the field of healthcare. Individual patient data can be inundative, but its value can be extracted by ML's predictive power and ability to find trends. A great area of interest is early diagnosis and disease management strategies for cardiovascular disease (CVD), the leading cause of death in the world. Treatment is often inhibited by analysis delays, but rapid testing and determination can help improve frequency for real time monitoring. In this research, an ML algorithm was developed in conjunction with a flexible BNP sensor to create a quick diagnostic tool. The sensor was fabricated as an ion-selective field effect transistor (ISFET) in order to be able to quickly gather large amounts of electrical data from a sample. Artifical samples were tested to characterize the sensors using linear sweep voltammetry, and the resulting data was utilized as the initial training set for the ML algorithm, an implementation of quadratic discriminant analysis (QDA) written in MATLAB. Human blood serum samples from 30 University of Pittsburgh Medical Center (UPMC) patients were tested to evaluate the effective sorting power of the algorithm, yielding 95% power in addition to ultra fast data collection and determination.