AJR
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 QUICK SEARCH:   [advanced]


     


This Article
Right arrow Abstract Freely available
Right arrow Full Text
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Eng, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Eng, J.
Social Bookmarking
 Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?

Predicting the Presence of Acute Pulmonary Embolism: A Comparative Analysis of the Artificial Neural Network, Logistic Regression, and Threshold Models

John Eng1

1 Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University, Central Radiology Viewing Area, Rm. 117, 600 N. Wolfe St., Baltimore, MD 21287.



View larger version (73K):

[in a new window]
 
Fig. 1. Schematic diagram illustrates structure of artificial neural network chosen for this study. Each lung was divided into three regions (upper, middle, and lower), resulting in 18 input units. Additional units were number of subsegmental perfusion mismatches, presence or absence of symmetry in the number of mismatches, and size of largest pleural perfusion. Links were defined for every possible pairing of input units between adjacent layers. Vent = ventilation scan, perf = perfusion scan, CXR = chest radiograph, Asymm = asymmetry.

 


View larger version (20K):

[in a new window]
 
Fig. 2. Graph depicts empiric (nonparametric) receiver operating characteristic curves representing accuracy of three data models in predicting presence or absence of pulmonary embolism. Curves are similar for each. Thick black line = neural network, thick gray line = logistic regression, thin black line = threshold.

 


View larger version (14K):

[in a new window]
 
Fig. 3. Graph shows comparison of areas under receiver operating characteristic (ROC) curves for four data models and two groups of physicians who were unaided by any modeling. Only diagnosis of PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) study's referring clinician was statistically significantly different from neural network model. SD = standard deviation, CI = confidence interval.

 


View larger version (17K):

[in a new window]
 
Fig. 4. Bar graph illustrates mean value of 21 input variables used in this study. Variable 19, representing total number of subsegmental ventilation—perfusion scanning mismatches, is dominant. {square} = pulmonary embolism absent, {blacksquare} = pulmonary embolism present.

 

Add to CiteULike CiteULike   Add to Complore Complore   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Digg Digg   Add to Reddit Reddit   Add to Technorati Technorati    What's this?




HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Copyright © 2002 by the American Roentgen Ray Society.