![]() |
|||
![]()
|
![]() |
![]() |
![]() |
6.2. SYSTEM INTEGRATIONMost of the expert systems that have been developed in the field of medicine are essentially stand-alone systems. However, in the very near future, it is likely that a large segment of the expert systems being developed in this field will be integrated with other information systems. Integration with a hospital information system permits the expert system to have direct access to the medical database. Such access is often of considerable value. Integrating a medical database with an expert system can greatly improve the knowledge base construction. This has been a major impediment to the development of expert systems, because such methods as neural network or case-based reasoning can acquire knowledge directly from the database through the interface mechanism. As the technology of the electronic medical record (EMR) improves, such integration will solve many technical problems in knowledge acquisition. Integration EMR can also greatly improve an accuracy of Bayesian system. Among the most commonly recognized problems with the use of a Bayesian approach has been the large amount of data required to determine all the conditional probabilities needed in the rigorous application of the formula. Such integration with the large medical database allows most of the necessary conditional probabilities to be obtained. Another form of integration is that of the interface between expert systems with medical instruments such as ECG, EEG, laboratory analyzers, CT scanner, and MRI. For example, while the EEG analysis system developed by Davey et al. (1989) can automatically detect spikes and sharp waves in the EEG for the detection of epileptiform activity, this system is stand-alone. On the other hand, a computerized ECG interpretation system capable of diagnosing complex cardiac arrhythmias (Tong et al., 1993) would be of great practical value in patient care by decreasing the time required from the ECG acquisition to physician interpretation because the expert system interfaces with the ECG online. The costs of health care would also decrease with such a system by reducing the need fro hujman expertise to interpret ECG recordings. In addition, the quality of patient care may increase as a result of the consistency of interpretations offered by the computerized analysis system. Similarly, the interpretation of CT scans can be automated by interfacing the expert system with CT scan (Natarajan et al., 1991). Since most of these medical instruments have an internal computer, such integration can easily be accomplished with current interfacing technology. Moreover, integration of fmedical instruments can also solve the problem of selecting proper domains because medical instruments are indispensable to the practice of medicine and physicians cannot perform adequately without them. 6.3. SYSTEM EVALUATIONWhile formal evaluation of medical expert system is acknowledged as an important requirement for acceptance of expert systems in clinical environment, the lack of appropriate evaluation studies contributes to clinicians' reluctance to medical expert systems. As a part of evaluation, the validation of the expert system performance encounters several major problems (O'Keefe et al., 1987). Decisions about what to validate, what to validate against, what to validate with, or when to validate find not ready-made guidelines to support them. Especially, validation of expert system become more problematic when no standards are available, as it is the case in a wide variety of clinical settings. The majority of validation studies solve such a situation by considering, in rather different ways, the opinion of several experts. A major criticism to this approach comes from the fact that different specialists do not always reach an agreement, and the opinion of several experts may not always be accurate. Martin-Baranera et al. (1996) suggested to introduce an external reference, through simulation of non-experts, in validating the expert system for assessing the etiology of community-acquired pheumonia in absence of standards. Diagnostic systems have tended to be assessed on the basis of their decision-making accuracy. Yet there are several additional components to the evaluation process when it is performed optimally. In order to demonstrate the effectiveness of a medical expert system, Shortliffe and Davis (1975) have suggested that system evaluations should be undertaken in a series of steps as follows:
Even though nearly 2 decades have passed since these seven-step evaluation criteria were published, not many medical expert systems have shown to meet formal validation criteria at all seven steps. In fact, most systems still have been assessed only at step 2, and remarkably few have met even the criterion of the need specified in step 1. However, if the system is to be used in a clinical environment rather than a research environment, formal system evaluations should be undertaken in these steps.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |