Fuzzy Support for Serodiagnosis:The ONSET Program


F Steimann, M Hayde, B Panzenböck, KP Adlassnig, A Pollak IEEE Engineering in Medicine and Biology Magazine 13 (1994) 705-709 Impact Factor: 0.427 (1994)


Abstract: General screening of pregnant women aims at the detection of acute primary maternal infection with Toxoplasma gondii and other pathogens threatening the life of the unborn. Fetal risk is directly dependent on the time of onset of infection relative to the date of conception: infections after conception must be treated, while those before need not. In current diagnosis, the expertise of a clinician is still required to assess the acuteness of infection. We developed a computer program called onset designed to explicate the clinician's knowledge and reasoning strategy by employing fuzzy representations of prototypical courses learned from evident acute infections. When presented with the serological findings of a particular case, it derives all possible times of onset of infection. In a first retrospective study, onset achieved conformity with the clinician's diagnosis in 91.4% of 1000 cases.

Keywords: fuzzy trend detection, serology, toxoplasmosis, onset of infection, supervised learning

paper