AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Application and Testing of Modifications to the TREAT Sepsis Network

Author

Term

4. term

Publication year

2012

Abstract

Sepsis is difficult to diagnose and treat, and early appropriate antibiotic therapy is critical. This thesis applies and tests modifications to the sepsis component of TREAT, a causal probabilistic network, to improve infection identification and antibiotic recommendations. Drawing on a targeted literature review and prior work, the study introduces structural changes (adding and redesigning nodes and revising state definitions) and replaces selected discrete chance nodes with continuous distributions. It also investigates automatic parameter learning using expectation–maximization in the Hugin environment, detailing the learning process, data preparation, mapping, and database requirements. The work outlines strategies to differentiate bacterial from viral infections and sepsis from non‑infectious SIRS. While implementation and testing of the changes are described, specific quantitative outcomes are not provided in this excerpt.

Sepsis er vanskelig at diagnosticere og behandle, og tidlig korrekt antibiotikabehandling er afgørende. Denne afhandling anvender og tester ændringer i sepsisdelen af TREAT, et kausalt probabilistisk netværk, for at forbedre identifikation af infektion og anbefaling af antibiotika. Med udgangspunkt i målrettet litteratursøgning og tidligere arbejde introduceres strukturelle ændringer (tilføjelse og redesign af noder samt reviderede tilstande) og udskiftning af udvalgte diskrete chancenoder med kontinuerte fordelinger. Der undersøges desuden automatisk parameterlæring med expectation–maximization i Hugin, inklusive beskrivelse af læringsprocessen, datatilberedning, mapping og databasekrav. Arbejdet skitserer strategier til at skelne mellem bakterielle og virale infektioner samt mellem sepsis og ikke‑infektiøs SIRS. Selvom implementering og test af ændringerne præsenteres, indeholder dette uddrag ingen specifikke kvantitative resultater.

[This apstract has been generated with the help of AI directly from the project full text]