Document Detail


Non-invasive classification of severe sepsis and systemic inflammatory response syndrome using a nonlinear support vector machine: a preliminary study.
MedLine Citation:
PMID:  20453293     Owner:  NLM     Status:  MEDLINE    
Abstract/OtherAbstract:
Sepsis has been defined as the systemic response to infection in critically ill patients, with severe sepsis and septic shock representing increasingly severe stages of the same disease. Based on the non-invasive cardiovascular spectrum analysis, this paper presents a pilot study on the potential use of the nonlinear support vector machine (SVM) in the classification of the sepsis continuum into severe sepsis and systemic inflammatory response syndrome (SIRS) groups. 28 consecutive eligible patients attending the emergency department with presumptive diagnoses of sepsis syndrome have participated in this study. Through principal component analysis (PCA), the first three principal components were used to construct the SVM feature space. The SVM classifier with a fourth-order polynomial kernel was found to have a better overall performance compared with the other SVM classifiers, showing the following classification results: sensitivity = 94.44%, specificity = 62.50%, positive predictive value = 85.00%, negative predictive value = 83.33% and accuracy = 84.62%. Our classification results suggested that the combinatory use of cardiovascular spectrum analysis and the proposed SVM classification of autonomic neural activity is a potentially useful clinical tool to classify the sepsis continuum into two distinct pathological groups of varying sepsis severity.
Authors:
Collin H H Tang; Paul M Middleton; Andrey V Savkin; Gregory S H Chan; Sarah Bishop; Nigel H Lovell
Related Documents :
22461853 - Cushing's syndrome and big igf-ii associated hypoglycaemia in a patient with adrenocort...
3754943 - The plica syndrome: a new perspective.
19782333 - Post-cardiac arrest syndrome: epidemiology, pathophysiology, treatment, and prognostica...
Publication Detail:
Type:  Journal Article; Research Support, Non-U.S. Gov't     Date:  2010-05-07
Journal Detail:
Title:  Physiological measurement     Volume:  31     ISSN:  1361-6579     ISO Abbreviation:  Physiol Meas     Publication Date:  2010 Jun 
Date Detail:
Created Date:  2010-05-21     Completed Date:  2010-09-22     Revised Date:  -    
Medline Journal Info:
Nlm Unique ID:  9306921     Medline TA:  Physiol Meas     Country:  England    
Other Details:
Languages:  eng     Pagination:  775-93     Citation Subset:  IM    
Affiliation:
School of Electrical Engineering and Telecommunications, The University of New South Wales, Sydney, NSW 2052, Australia.
Export Citation:
APA/MLA Format     Download EndNote     Download BibTex
MeSH Terms
Descriptor/Qualifier:
Adult
Algorithms
Analysis of Variance
Artificial Intelligence*
Autonomic Nervous System / physiopathology
Humans
Nonlinear Dynamics*
ROC Curve
Risk Factors
Sepsis / classification*,  diagnosis,  physiopathology

From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine


Previous Document:  Imaging hemorrhagic stroke with magnetic induction tomography: realistic simulation and evaluation.
Next Document:  Long-range correlations of different EEG derivations in rats: sleep stage-dependent generators may p...