Sepsis Detection Exploiting Biomarker Analysis with Deep Neural Networks

Machine learning has found widespread application in many domains, including medical sciences, offering promising solutions to complex healthcare challenges using cyber-physical systems (CPSs). Although mortality rates for sepsis and septic shock have declined in recent years, these conditions remain critical and are often overlooked, necessitating advanced predictive modeling to enable prompt and accurate diagnosis. Providing informative indicators for physicians and enabling early sepsis diagnosis are critical for improving patient outcomes, even when minimal information is accessible. In this paper, we present a deep learning (DL) pipeline for early sepsis detection when only a limited number of biomarkers is available. Furthermore,we also introduce a domain-guided data refinement pipeline that combines prior domain knowledge with the predictions of a DL model to guide the data gathering and refinement process. We demonstrate the effectiveness of the proposed approach through extensive experiments across a wide range of settings and tasks. Our findings show the potential of deep neural networks as a helpful tool for sepsis diagnosis, enabling timely interventions and better patient care.