Title :
Mixed neural-conventional processing to differentiate airway diseases by means of functional noninvasive tests
Author :
Parvis, Marco ; Gulotta, Carlo ; Torchio, Roberto
Author_Institution :
Dipt. di Electron., Torino Univ., Italy
fDate :
6/1/2001 12:00:00 AM
Abstract :
This paper describes a processing technique that can be used to combine information from different medical analyses to discriminate between different pathologies that have similar symptoms. The paper is focused on the differentiation between asthma, bronchitis, and emphysema, using only functional noninvasive tests, but the proposed technique can be easily applied to other similar situations where different tests have to be used to identify a pathology. The technique is based on mixed neural-and-conventional processing that not only suggests the pathology, but also estimates the reliability of this suggestion
Keywords :
Bayes methods; backpropagation; covariance matrices; diseases; error statistics; lung; medical diagnostic computing; multilayer perceptrons; pneumodynamics; Bayesian approach; airway diseases differentiation; asthma; backpropagation; bronchitis; competitive layer; covariance matrix; emphysema; error probability; functional noninvasive tests; linear discriminant score; lung pathologies discrimination; mixed neural-conventional processing; multilayer perceptron; pathology evidence index; spirometric data; uncertainty; Diseases; Information analysis; Lungs; Medical diagnostic imaging; Medical tests; Neural networks; Pathology; Patient monitoring; Testing; Uncertainty;
Journal_Title :
Instrumentation and Measurement, IEEE Transactions on
