Predictors of urinary tract infection based on artificial neural networks and genetic algorithms

Paul S. Heckerling, Gay J. Canaris, Stephen D. Flach, Thomas Gerald Tape, Robert Swift Wigton, Ben S. Gerber

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

Background: Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. Methods: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of ≥105 colony-forming units (CFU) per milliliter, and counts of ≥102 CFU per milliliter. Results: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of ≥105 CFU per milliliter) to 0.792 (for uropathogen counts of ≥102 CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. Conclusions: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.

Original languageEnglish (US)
Pages (from-to)289-296
Number of pages8
JournalInternational Journal of Medical Informatics
Volume76
Issue number4
DOIs
StatePublished - Apr 1 2007

Fingerprint

Urinary Tract Infections
Urinalysis
Stem Cells
Urine
Dysuria
Infection
ROC Curve
Erythrocytes
Epithelial Cells
Bacteria

Keywords

  • Artificial neural networks
  • Genetic algorithms
  • Urinary tract infection

ASJC Scopus subject areas

  • Health Informatics

Cite this

Predictors of urinary tract infection based on artificial neural networks and genetic algorithms. / Heckerling, Paul S.; Canaris, Gay J.; Flach, Stephen D.; Tape, Thomas Gerald; Wigton, Robert Swift; Gerber, Ben S.

In: International Journal of Medical Informatics, Vol. 76, No. 4, 01.04.2007, p. 289-296.

Research output: Contribution to journalArticle

@article{91441e86f3d84232a84d0f66471b7f85,
title = "Predictors of urinary tract infection based on artificial neural networks and genetic algorithms",
abstract = "Background: Among women who present with urinary complaints, only 50{\%} are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. Methods: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of ≥105 colony-forming units (CFU) per milliliter, and counts of ≥102 CFU per milliliter. Results: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of ≥105 CFU per milliliter) to 0.792 (for uropathogen counts of ≥102 CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. Conclusions: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.",
keywords = "Artificial neural networks, Genetic algorithms, Urinary tract infection",
author = "Heckerling, {Paul S.} and Canaris, {Gay J.} and Flach, {Stephen D.} and Tape, {Thomas Gerald} and Wigton, {Robert Swift} and Gerber, {Ben S.}",
year = "2007",
month = "4",
day = "1",
doi = "10.1016/j.ijmedinf.2006.01.005",
language = "English (US)",
volume = "76",
pages = "289--296",
journal = "International Journal of Medical Informatics",
issn = "1386-5056",
publisher = "Elsevier Ireland Ltd",
number = "4",

}

TY - JOUR

T1 - Predictors of urinary tract infection based on artificial neural networks and genetic algorithms

AU - Heckerling, Paul S.

AU - Canaris, Gay J.

AU - Flach, Stephen D.

AU - Tape, Thomas Gerald

AU - Wigton, Robert Swift

AU - Gerber, Ben S.

PY - 2007/4/1

Y1 - 2007/4/1

N2 - Background: Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. Methods: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of ≥105 colony-forming units (CFU) per milliliter, and counts of ≥102 CFU per milliliter. Results: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of ≥105 CFU per milliliter) to 0.792 (for uropathogen counts of ≥102 CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. Conclusions: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.

AB - Background: Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. Methods: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of ≥105 colony-forming units (CFU) per milliliter, and counts of ≥102 CFU per milliliter. Results: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of ≥105 CFU per milliliter) to 0.792 (for uropathogen counts of ≥102 CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. Conclusions: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.

KW - Artificial neural networks

KW - Genetic algorithms

KW - Urinary tract infection

UR - http://www.scopus.com/inward/record.url?scp=33847257363&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33847257363&partnerID=8YFLogxK

U2 - 10.1016/j.ijmedinf.2006.01.005

DO - 10.1016/j.ijmedinf.2006.01.005

M3 - Article

C2 - 16469531

AN - SCOPUS:33847257363

VL - 76

SP - 289

EP - 296

JO - International Journal of Medical Informatics

JF - International Journal of Medical Informatics

SN - 1386-5056

IS - 4

ER -