Bosss And Iliad

As indicated in the previous discussion, researchers have developed many diagnostic expert systems that vary in both the underlying technology as well as the domain of application. In this section, we briefly describe two systems that illustrate the key characteristics of probabilistic diagnostic expert systems. These systems—Iliad and BOSSS—are diagnostic

figure 13.1 The process of diagnosis. Like a physician, a diagnostic expert system analyzes patient data to generate a differential diagnosis. Based on the differential diagnosis and value of information calculations, the diagnostic expert system can suggest additional questions to ask the patient, additional physical observations to make about the patient, or additional tests to consider. A physician may accept the recommendations or not. As additional information about the patient's illness arrives over time (either as a result of tests or questions suggested by the computer; or by the physician; or simply the passage of time, this user can enter this new information and rerun the computer program to generate an updated differential diagnosis). This process of data collection and analysis can be repeated frequently until the diagnosis for a patient is established with sufficient certainty and diagnostic precision that further diagnostic evaluation is not necessary.

figure 13.1 The process of diagnosis. Like a physician, a diagnostic expert system analyzes patient data to generate a differential diagnosis. Based on the differential diagnosis and value of information calculations, the diagnostic expert system can suggest additional questions to ask the patient, additional physical observations to make about the patient, or additional tests to consider. A physician may accept the recommendations or not. As additional information about the patient's illness arrives over time (either as a result of tests or questions suggested by the computer; or by the physician; or simply the passage of time, this user can enter this new information and rerun the computer program to generate an updated differential diagnosis). This process of data collection and analysis can be repeated frequently until the diagnosis for a patient is established with sufficient certainty and diagnostic precision that further diagnostic evaluation is not necessary.

expert systems in the domains of diseases of humans and cattle, respectively.

4.1. Iliad

Iliad was developed by Homer R.Warner, Jr. and colleagues at the University of Utah in the 1980s (Warner, 1989, Sorenson et al., 1988, Cundick et al., 1989, Warner and Bouhaddou, 1994, Warner, 1990). Iliad is a stand-alone diagnostic expert system; that is, it requires a user (physician) to enter findings into the program. Iliad uses Bayes rules to compute a differential diagnosis for the patient's illness. It uses value-of-information calculations to recommend additional tests and observations.

Iliad performs differential diagnosis of human diseases in a variety of fields: internal medicine, sports medicine, pediatrics, dermatology, psychiatry, obstetrics/gynecology, urology, peripheral vascular diseases, and sleep disorders. Iliad covers more than 900 diseases and 1500 syndromes (which means that there are tables of information about diseases and findings for approximately 2400 diseases similar to those shown in Figures 13.1 and 13.2). Iliad computes a differential diagnosis for a patient and then displays the diseases and syndromes to the user in order of posterior probability as shown in Figure 13.2. If the user asks Iliad to suggest additional tests or questions to ask the patient, the program performs value-of-information calculations to suggest findings that would discriminate among the diseases in the differential diagnosis (Figure 13.3).

Researchers have compared Iliad's diagnostic performance on real cases with that of expert physicians and found equivalent performance (Graber and VanScoy, 2003, Friedman et al.,

0 0

Post a comment