Display Settings:

Format

Send to:

Choose Destination
See comment in PubMed Commons below
World Neurosurg. 2011 Jan;75(1):57-63; discussion 25-8. doi: 10.1016/j.wneu.2010.07.007.

Prediction of symptomatic cerebral vasospasm after aneurysmal subarachnoid hemorrhage with an artificial neural network: feasibility and comparison with logistic regression models.

Author information

  • 1From the Division of Neurosurgery, Departmentof Surgery, University of Vermont, Burlington,Vermont, USA. Travis.Dumont@vtmednet.org

Abstract

OBJECTIVE:

To create a simple artificial neural network (ANN) to predict the occurrence of symptomatic cerebral vasospasm (SCV) after aneurysmal subarachnoid hemorrhage (aSAH) based on clinical and radiographic factors and test its predictive ability against existing multiple logistic regression (MLR) models.

METHODS:

A retrospective database of patients admitted to a single academic medical center with confirmed aSAH between January 2002 and January 2007 (91 patients) was input to a back-propagation ANN program freely available to academicians on the Internet. The resulting ANN was prospectively tested against two previously published MLR prediction models for all patients admitted the following year (22 patients). The models were compared for their predictive accuracy with receiver operating characteristic (ROC) curve analysis.

RESULTS:

All models were accurate with their prediction of patients with SCV. The ANN had superior predictive value compared with the MLR models, with a significantly improved area under ROC curve (0.960 ± 0.044 vs 0.933 ± 0.54 and 0.897 ± 0.069 for MLR models).

CONCLUSIONS:

A simple ANN model was more sensitive and specific than MLR models in prediction of SCV in patients with aSAH. The conception of ANN modeling for cerebral vasospasm is introduced for a neurosurgical audience. With advanced ANN modeling, the clinician may expect to build improved models with more powerful prediction capabilities.

Copyright © 2011 Elsevier Inc. All rights reserved.

Comment in

PMID:
21492664
[PubMed - indexed for MEDLINE]
PubMed Commons home

PubMed Commons

0 comments
How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Elsevier Science
    Loading ...
    Write to the Help Desk