MYCIN

From Clinfowiki
Jump to: navigation, search

MYCIN was a clinical decision support system developed in 1972 at Stanford University by Edward Shortliffe as a consultation system that focused on appropriate management of patients who had infections. The system was never publicly used in clinical practice settings. MYCIN is not an acronym. The name was chosen after attempts at finding a suitable acronym failed. The name is the common suffix associated with many antimicrobial agents. [1] In studies, MYCIN gave advice that was compared favorably with that offered by experts in infectious diseases. MYCIN is not currently being used. [2]

Background

MYCIN was an outcome learned in the construction of DENDRAL, the name of the group of several large programs which helped organic chemists explain molecular structure through specialized knowledge of chemistry. DENDRAL program was the first artificial intelligence (AI) program to emphasize specialized knowledge over generalized problem solving methods. [3] MYCIN used roughly 600 production rules to determine the infectious disease diagnosis. Production rule is a conditional statement that relates observations to associated inferences that can be drawn. MYCIN would take the physician’s replies to the questions into consideration and provide a list of possible disease diagnosis ranked from high to low based on the probability of each. The first grant application for MYCIN was submitted in October 1973. Mycin is an expert system in that it is an artificial intelligence program designed (a) to provide expert-level solutions to complex problems, (b) to provide interactive explanation capabilities, and (c) to be flexible enough to accommodate new judgement knowledge easily. [4] There are two main parts to an expert system like MYCIN: a knowledge base and an inference mechanism, or engine. The MYCIN system is designed to be a problem-solving to aid in decision making using an algorithm to provide a solution to the problem. [1] MYCIN is an evidence gathering program that uses back chaining or goal directed reasoning. It starts with a stated goal to achieve and work backwards through inference rules from right to left to find the data that establishes the goal. [2]

Statistics

Research conducted at Stanford Medical School on therapy selection for patients with meningitis found that MYCIN gave an acceptable therapy in about 69% of the cases, which compared favorably with that offered by experts in infectious diseases. MYCIN was a factor in improving research and.

Clinical Applications

Lack of technology during its development in 1972 led to few clinical uses of MYCIN. LISP, the second oldest high level programming language, was chosen for programming because it was an interpretive language that allowed rapid modification and testing and did not require programs be recompiled after modification. LISP was flexible in allowing the use of rules as part of the MYCIN program. [5] MYCIN required all relevant data about a patient to be entered by typing the responses into a stand alone system. This took more than 30 minutes for a physician to enter in the data, which was an unrealistic time commitment. In current times, an EHR would extract answers to questions from a patient’s database.

References

  1. 1.0 1.1 http://www.amia.org/staff/eshortliffe/Buchanan-Shortliffe-1984/Chapter-01.pdf
  2. 2.0 2.1 Musen, M., Shahar, Y., & Shortliffe, E. (2006). Clinical decision-support systems. In E. Shortliffe & J. Cimino (Eds.), Biomedical informatics: Computer applications in health care and biomedicine (pp. 698 – 736). New York, NY: Springer.
  3. Buchanan, B. G., & Shortliffe, E. H. (1984). Rule-based expert systems: The MYCIN experiments of the Stanford Heuristic Programming. Reading, MA: Addison-Wesley (p. 8)
  4. Buchanan, B. G., & Shortliffe, E. H. (1984). Rule-based expert systems: The MYCIN experiments of the Stanford Heuristic Programming. Reading, MA: Addison-Wesley (p. 10)
  5. Buchanan, B. G., & Shortliffe, E. H. (1984). Rule-based expert systems: The MYCIN experiments of the Stanford Heuristic Programming. Reading, MA: Addison-Wesley (p. 6)