Generation of an intelligent medical system, using a real database, to diagnose bacterial infection in hospitalized patients.

The initial diagnosis of bacterial infections in the absence of laboratory microbiological data requires physicians to use clinical algorithms based on symptoms, patient history and infection site. Optimization of such algorithms would be achieved by including as many variables associated with bacterial infection as possible. Demographic data are easily available and frequently used to sub-group human populations. A prospective investigation was, therefore, undertaken to examine the influence of demographic variables on bacterial infection rates, using data obtained from 173 patients presenting to Albert Einstein Medical Center. Data was randomly selected from 149 of these patients and used to generate fuzzy rules to model an intelligent medical system. To test the accuracy of this system at determining bacterial infection, based solely on demographic data, the program was given the remaining 24 patients' information. All 18 patients with either streptococcal, staphylococcal or Escherichia coli infections were correctly diagnosed. Non-E.coli GNR were misdiagnosed as E. coli infections in two patients resulting in an overall prediction rate for the 24 patients of 91.66%. This study suggests that the direct correlation of demographic variables with a predisposition to bacterial infection allow the design of an intelligent medical system, which shows great future potential as a diagnostic tool for all physicians.

Using fuzzy sets to analyze putative correlates between age, blood type, gender and/or race with bacterial infection.

Previous studies have suggested that the demographic variables of age and blood type may serve as "risk factors" for infection by specific bacterial species. Since both demographic variables and bacterial species are defined using generally accepted parameters, they constitute highly suitable variables for the generation of a fuzzy logic program. A prospective study was therefore undertaken to examine the influence of age, blood type, gender and race on bacterial infection rates using a real database generated from 187 bacteremic patients admitted to Albert Einstein Medical Center. A fuzzy logic program was created using 155 randomly selected patients' data with four input (demographic variables) and four output classes (infections with "staphylococci", "streptococci", "Escherichia coli" or "non-E. coli gram negative rods (non-E.coli GNR)"). To see whether bacterial infection could be predicted based on demographic data alone, the program was tested using the remaining 32 patients' data. The program was able to correctly determine the bacterial output group of 27 of 32 randomly selected patients, giving an overall correlation of 84.38%. This study suggests that the direct correlation of demographic variables with a predisposition to bacterial infection allow the design of an intelligent medical system, which shows great future potential as a powerful diagnostic tool for all physicians.

A neuro-fuzzy algorithm for diagnosis of coronary artery stenosis.

In this paper a method of fuzzy decision making applied to diagnosis of coronary artery stenosis is presented. The method uses a neural network approach for the diagnosis of stenosis in the three main coronary arteries (left anterior descending, right coronary artery, and circumflex). First, the knowledge base domain, 201Tl scintigram training data, is explained and the method of preprocessing the original heart images is given. Next, the method of dealing with the uncertainties present in the data using the fuzzy approach is outlined. Finally, the algorithm and the results are discussed and compared with other approaches.