Development of an early diagnostic system using fuzzy theory for postoperative infections in patients with gastric cancer

Dig Surg. 2004;21(3):210-4. doi: 10.1159/000079394. Epub 2004 Jun 25.

Abstract

Background: Early prediction of infection following surgery for gastric cancer may permit earlier intervention.

Aim: The aim was to develop an early diagnostic system for postoperative infection.

Methods: Clinical and laboratory data were analyzed in 180 patients who had surgery for gastric cancer at the Wakayama Medical University Hospital (January 1992 to December 1994). Of these, 60 patients developed a postoperative infection. A predictive system was then devised using fuzzy theory and evaluated in a second set of 137 patients who underwent surgery for gastric cancer at Wakayama Medical University and 10 other associated hospitals (August 1995 to January 1997).

Results: The system identified seven parameters grouped into six rules and entered into a fuzzy logic system to predict either infection or non-infection. (1) Blood loss, and extent of resection; (2) the febrile pattern during days 2 to 4; (3) differential leukocyte count on day 4; (4) C-reactive protein level on day 4; (5) time sequential vectors during days 1-4 for the band cell neutrophil count, and (6) leukocyte count. The sensitivity was 86% (24/28), the specificity was 90% (98/109), the negative predictive value was 96% (98/102), the positive predictive value was 69% (24/35) and the overall accuracy was 80% (122/137).

Conclusion: Fuzzy theory can provide early prediction of postoperative infection using standard clinical and biochemical parameters. The clinical utility of this system needs to be determined in future studies.

MeSH terms

  • Diagnosis, Computer-Assisted*
  • Female
  • Fuzzy Logic*
  • Gastrectomy
  • Humans
  • Infections / diagnosis*
  • Male
  • Middle Aged
  • Postoperative Complications / diagnosis*
  • Predictive Value of Tests
  • Sensitivity and Specificity
  • Stomach Neoplasms / surgery*