The Sleep-EVAL Expert System

First created | 01/12/1994

Last edited | 05/11/2012

Summary by Maurice M. Ohayon, MD, DSc, PhD

Reference to cite: Ohayon M. Validation of expert systems: Examples and considerations. Medinfo 1995; 8:1071-5.

Sleep-EVAL, an artificial intelligent computer program, is an Expert System for evaluation and diagnosis of Sleep and Mental Disorders in general and clinical populations

What is an expert system?

An expert system is a computer program conceived to simulate some forms of human reasoning (by the intermediary of an inference engine) and capable to manage an important quantity of specialized knowledge.

These capacities for reasoning and management allow the system to target a small number of relevant hypotheses in the mass of potential diagnoses and being able to find a satisfactory diagnostic conclusion.

Two characteristics of the expert system are essential to accomplish this task:

the aptitude to process an important mass of specialized knowledge and
the aptitude to simulate the human reasoning (in an imperfect manner).

Interests and limits

There are many well known advantages to using computerized tools and expert systems:

reduction of missing data,
better collection of data,
no omission of questions,
no data transcription,
broader coverage of diagnoses, etc...

Even if some computerized tools possess some diagnostic trees, the term “computerized tools” is not synonymous with “expert systems”.

Over their apparent similarities, they are radically different in terms of both conception and capabilities.

Indeed, the apparent reasoning process in computerized tools is only an artifice: The diagnostic trees are predetermined and the software only goes from one node to another without attempting to look for other paths.

Expert systems are making their decision during the interview, looking for the optimal way to reach their conclusions: to make a diagnosis.

Obviously, some limits remain.

Both types of instruments do not possess the richness of the human language and some may complain of a inflexibility in wording.

They cannot analyze non-verbal information such as a lack of hygiene, etc., nor detect a contradiction between a verbal answer and behavioral cues.

Most computerized tools are unable to analyze temporal information, for example, to determine which symptom appeared first unless, like Sleep-EVAL, they have a mathematical preprocessor able to make this type of analysis.

Conclusion

There are two ways to proceed in order to improve epidemiological studies using an expert system like Sleep-EVAL:

- The first lies in the improvement of the questionnaire. This can be accomplished by increasing the quality of data collected by allowing fuzzy and uncertain answers.

- The second way introduces a sharper knowledge of the elements belonging to a diagnosis. This can be done by calculating the relative weight of those elements in the diagnosis of a given pathology. Quality, frequency, intensity are all elements that provide indication on the strength of the associative links between the different criteria and the diagnosis.

References

Ohayon MM. Improving decision making processes with the fuzzy logic approach in the epidemiology of sleep disorders. J Psychosom Res 1999;47:297-311.

Ohayon MM, Guilleminault C, Zulley J, Palombini L, Raab H. Validation of the Sleep-EVAL system against clinical assessments of sleep disorders and polysomnographic data. Sleep 1999; 22:925-30.

Ohayon M. Validation of expert systems: Examples and considerations. Medinfo 1995; 8:1071-5.

Ohayon M. Knowledge Based System Sleep-EVAL: Decisional Trees and Questionnaires. Bibliothèque Nationale du Québec, Bibliothèque Nationale du Canada, ISBN 2-921483-06-8, 1995.

Ohayon MM. Sleep-EVAL, Knowledge Based System for the Diagnosis of Sleep and Mental Disorders. Registration #437699, Copyright Office, Canadian Intellectual Property Office. Ottawa: Industry Canada, 1994. (English, French, German, Italian, Portuguese, and Spanish versions).

Ohayon M. [Expert systems in psychiatry: current nosographic orientation]. Ann Med-Psychol (Paris) 1987;145(6):521-526.

Ohayon M. [From logical data access to the development of expert systems] Psychol Med (Paris) 1986;18(4):578-580.

Ohayon M. [Data capture and computer sciences: interest in the assessment of psychotropic medication] Psychol Med (Paris) 1986;18(4):581-583.

Ohayon M, Fondaraï, J. [Similarities and differences between DSM III and French psychiatric practice]. Ann Med-Psychol (Paris) 1986;144(5):515-530.

More Information

Sleep-EVAL Aims

When the first epidemiological survey was launched in 1992, we were looking for an assessment tool that could be used by interviewers with little knowledge about sleep disorders.

Knowledge Base
The inference engine uses its knowledge base to pose questions, to infer hypotheses and to deduce diagnostic conclusion

Inference Engine
Sleep-EVAL is a non-monotonic, level-2 expert system endowed with the ability to make logical connections based on patient information (causal reasoning mode).

Fuzzy Logic
Inference models such as probabilistic and fuzzy systems can be used to integrate uncertainty in both symptomatic assessment and diagnostic attribution. It therefore becomes possible to extend boundaries and attribute a degree of certainty to a diagnosis.
A probabilistic model can be easily computed from an existent binary data set.
A fuzzy model can also be calculated from an existent data set, but the model is obviously much more precise when the data are expressed in categorical terms.