A Multilevel Expert System in Medicine - CiteSeerX

12 downloads 17105 Views 249KB Size Report
in a format that can be represented internally in the computer memory; ... consultant to doctors in investigating patients for infections, blood bacterial matters and ...
The current issue and full text archive of this journal is available at www.ijasr. org/ ij as r-v11.ht ml

A Multilevel Expert System in Medicine Liviu Octavian Mafteiu-Scai

Expert Medicine System 1

National College “Iancu de Hunedoara”, Hunedoara, Romania Abstract- This paper presents a multilevel expert system architecture for the diagnosis, prognosis and monitoring in medicine. These problems are solved with procedural programming (C language) and logical programming (CLIPS language). This double approach allowed a comparison between different methods of implementing an expert system. An important element of this project is an expert system generator, evidenced by automatically generating of the source code in CLIPS language. As a result of this, the proposed expert system allows import/export of knowledge between users, without requiring the presen ce of the kno wled g e engin eer. Keywords- expert system, med icin e, exp ert system generat o r Paper type- Research paper

International Journal of Academic & Scientific Research Vol. 1, Issue 2 No. 2, 2013 pp. 1-6

IJASR 1, 2, 1

2

I. INTRODUCTION At least one patient of 12 who died were diagnosed incorrectly as mentioned in an analysis of The Journal of the American Medical Association ( Vol 289 No. 21, June 4, 2003), fact that generate a great interest in creating specific software tools to help the doctors in making a correct decision. Artificial intelligence has been created and developed to provide a solution to those unsolvable problems through algorithmic methods, using other methods that are close to the human reasoning. The oldest branch of AI, expert systems (ES) are software systems based on AI techniques, which store human expert knowledge from a clearly defined area and then use them to solve problems in this area. Expert systems are software applications that solves problems that would otherwise require extensive and complicated human expertise, by simulates the human reasoning process using specific knowledge and facts. The main elements of an expert system are [1]: - Knowledge base, contains all the specialized knowledge taken from human experts. Knowledge stored in the knowledge base reflects real world objects and the relationships between them; - Facts base, contains facts that describe the initial problem statement and the intermediate results produced during the inference process; - Rules base, contains all the rules applicable to the facts in order to obtain reasoning’s; - inference engine, the processing element that takes expert systems knowledge stored in the knowledge base and build reasoning’s needed to solve the problem; - Explanatory module serves to justify the reasoning made by the inference engine; - Knowledge acquisition module, transforms the knowledge taken from human expert in a format that can be represented internally in the computer memory; - User interface. From the point of view of their goals, expert systems can be classified as: a) Classification-interpretation ES: generate only a technical diagnosis. Expert systems from this category have the knowledge represented in the form of rules. Knowledge is empirical and therefore the system cannot justify the reasoning. ES from this class do an exhaustive search, which applies all possible rules to increase or decrease the certainty factor of the solution. b) Monitoring-control ES: the main factor is the time. These ES is for supervision and/or management processes. When certain conditions are satisfied in the knowledge base at a specific time, ES triggers a specific action that represent a future strategy. c) Prognosis-anticipation ES: are able to generate prognoses based on current information and evolutions from the past.

The most popular medical expert systems: -

MYCIN was one of first ES used in medicine [2]. This system was designed as a consultant to doctors in investigating patients for infections, blood bacterial matters and appropriate treatment. The system was able to explain the reasoning. Its architecture is based on backward chaining production system.

-

INTERNIST is an expert system developed at the University of Pittsburgh in 1982 and covers 80% of internal medicine [3]. Each disease is related to the torque observation-diagnosis;

-

HELP (Health Evaluation through Logical Processes) is a knowledge based information system in a hospital [4]. It has a decision function based on knowledge of radiology, documentation, research nurse, pharmacology and monitoring patients;

-

PEIRS (Pathology Expert Interpretative Reporting System) is an expert system that adds comments pathology to the generated reports [5]. Knowledge acquisition strategy allows pathologist to build over 2300 rules. The system includes thyroid function tests and glucose tolerance tests.

-

Puff is an expert system that analyzes lung function tests [6]. Some further implementation are still used.

II. THE PROPOSED SYSTEM The proposed system can perform all three functions: classification, monitoring and anticipation. The main functions of the proposed system are: -

classification: patient diagnosis and recommendation an appropriate treatment; monitoring: disease evolution monitoring by analyzing the parameters provided by monitoring/investigation medical devices (hardware devices); Determining the possible evolution of the disease. The basic functional structure of the proposed expert system is represented in Figure 1 and in Figures 2,3 and 4 are represented the basic architecture of the main sub-modules. The used inference engines operates in four tempos i.e.: Selection, filtering, solving conflicts and execution.

Expert Medicine System

3

IJASR 1, 2, 1

4

Figure 1: Functional structure

Figure 2: Acquisition module architecture

Figure 3: The structure of knowledge representation

Expert Medicine System 5

Figure 4: Knowledge processing module Other implemented functions in the current stage of system implementation are: -

-

-

-

Data transfer (knowledge base) between users. Function allows the knowledge base (symptoms, diseases and treatments) to be complemented with other knowledge from different users. Note that due to internal implementation, are concatenated only the knowledge of the user B that are not in the set of user A knowledge, thus avoiding data redundancy; In case of adding new symptoms/diseases in the knowledge base, it is not allowed the existence of two identical names for the same basic symptom/disease. Each basic symptom/disease has a unique identifier, automatically generated by the application; Diagnosis / prognosis / treatment is the most complex part of the proposed system. The mechanism used for these is forward chaining. Here, with techniques iterative (using C + +) and techniques from IA (using CLIPS), are determined all possible diseases for a set of facts entered by user; patient monitoring is done through personal medical records; CLIPS (C Language Integrated Production System) [7] source files are automatically generated by the application. This component fits into the category of "Expert System Generator", which gives a high degree of generality. Three CLIPS source code files are automatically generated: OR approach, AND approach and an approach where the outcome of the analysis depends on the answers that the user provides to the system. The architecture of Expert System Generator is represented in Figure 5.

IJASR 1, 2, 1

6

Figure 5: Expert system generator architecture

III. CONCLUSION The proposed multilevel expert system architecture can represent an improvement of the medical process, as a major help for the doctor regarding the diagnosis, treatment and monitoring of patients. The Expert System Generator, included in this project and evidenced by automatically generating of the source code in CLIPS language, is a way of increasing the adaptability of the system to new medical results without the intervention of a knowledge engineer. A future direction will be the inclusion of fuzzy logic engines, like Fuzzy CLIPS [8], in proposed expert system.

IV. REFERENCES [1] J.C. Giarratano, G.D. Riley, Expert System, Principles and Programming, fourth edition, ISBN-10: 0534384471, 2004 [2] Rule-Based Expert Systems -The MYCIN Experiments of the Stanford Heuristic Programming Project, B.G. Buchanan and H. Edward editors, Addison -Wesley, 1984 [3] Miller, R. & all, INTERNIST-1: An Experimental Computer-Based Diagnostic Consultant for General Internal Medicine.- New England Journal of Medicine 307, 1982 [4] Gardner RM, Pryor TA, Warner HR. The HELP hospital information system, International Journal of Medical Informatics 54, pp.169-182, 1999 [5] Edwards G, Compton P, Malor R, Srinivasan A, Lazarus L, PEIRS: a pathologistmaintained expert system for the interpretation of chemical pathology reports, Pathology 1993 Jan. 25(1), pp.27-34, 1993 [6] Janice S. Aikins,.John C. Kunz, Edward H. Shortliffc, and Kobcrt .J.Fallat, PUFF: An Expert System for Interpretation of Pulmonary Function Data, Report No. STANCS-82-931, September I982, Department of Chemical Pathology, St Vincent's Hospital, Sydney, 1982 [7] Culbert,C.,Riley,G.,Donnell,B., CLIPS Reference Manual,Vol.1-3. Johnson Space Center NASA, 1993 [8] R. Orchard, FuzzyCLIPS User’s guide, Integrating Reasoning Institute for Information Tech. National Research Council Canada, 1998