Observer Design and Analysis for Non-Invasive Hemorrhage Detection. Issue 20 (2021)

• Record Type: Journal Article
• Authors: Jin, Xin; Chalumuri, Yekanth Ram; Tivay, Ali; Hahn, Jin-Oh
• Is Part Of: IFAC-PapersOnLine. Volume 54:Issue 20(2021)
• Page Start: 310
• Page End: 315
• Publication Date: 2021
• Subjects: observer -- detection -- fault -- hemorrhage -- hemoglobin; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2021.11.192
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: This paper presents the design and analysis of an observer to enable non-invasive hemorrhage detection. The proposed observer-based approach detects hemorrhage by continuously receiving the volume resuscitation inputted to a patient and the resulting hemoglobin response of the patient, and creating multiple state estimates as candidate signatures. A unique challenge differentiating the hemoglobin-based hemorrhage detection problem from conventional fault detection problems is that hemorrhage (i.e., the process fault) also alters the measurement equation. Through the design of an observer using a blood volume kinetics model as plant dynamics, and the extensive analysis of its error dynamics incorporating the process fault-induced alteration in sensing, it was demonstrated that hemorrhage can be detected based on the dynamic behaviors of the signatures generated by the observer: virtual in silico testing resulted in a F-score of 0.80 (precision: 0.91; recall: 0.78) and a normalized detection time of 0.1-0.7 relative to the time to 25% blood volume loss.