0 Descision Support Systems In Medicine Shojaee

647 views

Published on

1 Comment
2 Likes
Statistics
Notes
No Downloads
Views
Total views
647
On SlideShare
0
From Embeds
0
Number of Embeds
14
Actions
Shares
0
Downloads
0
Comments
1
Likes
2
Embeds 0
No embeds

No notes for slide

0 Descision Support Systems In Medicine Shojaee

  1. 1. In the Name of<br />the promoter<br />of<br />Wisdom and Beauty<br />
  2. 2. By: Dr. Abbas Shojaee M.D., PhD stu.<br />An Introduction to Decision Support Systems in Medicine<br />This presentation uses works of:<br /><ul><li>Joseph Tan – Ehealth care information systems (book)
  3. 3. J.H. Van Bemmel - Medical informatics (book)
  4. 4. Edward H. Shortliffe – Biomedical Informatics (book)
  5. 5. Kate Farrell – Intro expert systems test (ppt file)</li></li></ul><li>Introduction<br />Every human action is the next step to data acquisition and interpretation.<br />Practicing medicine is making decisions indeed.<br />Computers are presumed that will help in these three steps and so decision Support in medicine.<br />Due to several limitations in storage, representation and interpretation of information, it is yet limited to well structured and actionable data and knowledge.<br />Questions exist:<br />What?<br />How?<br />When?<br /> At what quality?<br />How they affect the field?<br />
  6. 6. Decision making in medicine<br />Iterates<br />Interpretation<br />
  7. 7. Types of clinical decisions<br />Clinical decisions:<br />Diagnosis<br />What is true?<br />Diagnosis process.<br />Which information is needed to make right decision<br />Treatment<br />Which is the best pass way through the jungle of possibilities<br />Requirements:<br />Accurate and adequate data.<br />Applicable knowledge.<br />Appropriate problem solving skills. (inference engine)<br />
  8. 8. Some obstacles to use computers as DSS in medicine<br />Abundance of rules and interacting data.<br />Computational models are not as good as human inference models<br />Yet we do not know the model(s) that brain uses to infer.<br />Brain computational power is higher than our computers.<br />
  9. 9. Additional terms<br />Inductive and deductive reasoning<br />Classification<br />Feature set<br />Clustering<br />Decision model<br />Supervised and Unsupervised learning<br />Differential diagnosis<br />
  10. 10. Expert Systems<br />An expert system is a computer program that is designed to hold the accumulated knowledge of one or more domain experts<br />
  11. 11. Applications of Expert Systems<br />PUFF:<br />Medical system for diagnosis of respiratory conditions<br />PROSPECTOR:<br />Used by geologists to identify sites for drilling or mining<br />
  12. 12. Applications of Expert Systems<br />MYCIN:<br />Medical system for diagnosing blood disorders. First used in 1979<br />DESIGN ADVISOR:<br />Gives advice to designers of processor chips<br />
  13. 13. Applications of Expert Systems<br />DENDRAL: Used to identify the structure of chemical compounds. First used in 1965<br />LITHIAN: Gives advice to archaeologists examining stone tools<br />
  14. 14. Components of an Expert System<br />The knowledge base is the collection of facts and rules which describe all the knowledge about the problem domain<br />The inference engine is the part of the system that chooses which facts and rules to apply when trying to solve the user’s query<br />The user interface is the part of the system which takes in the user’s query in a readable form and passes it to the inference engine. It then displays the results to the user.<br />
  15. 15. Why use Expert Systems?<br />It takes long time and great costs to train an expert.<br />Experts are not always available. An expert system can be used anywhere, any time.<br />Human experts are not 100% reliable or consistent<br />Experts may not be good at explaining decisions<br />Cost effective<br />
  16. 16. Some Problems with Expert Systems<br />Limited domain<br />Systems are not always up to date, and don’t learn<br />No “common sense”<br />Experts needed to setup and maintain system<br />
  17. 17. Legal and Ethical Issues<br />Who is responsible if the advice is wrong?<br />The user?<br />The domain expert?<br />The knowledge engineer?<br />The programmer of the expert system shell?<br />The company selling the software?<br />
  18. 18. Medical Expert Systems<br />
  19. 19. Codifying Human Knowledge<br /><ul><li>Decomposition into “chunks” of knowledge, chaining of inferences
  20. 20. Matching of case data to prototypical situations
  21. 21. Using causal models (pathophysiology) to figure out cases</li></li></ul><li>Mycin: Rule-based Systems<br />Task: Diagnosis and prescription for bacterial infections of the blood (and later meningitis)<br />Method: <br />Collection of modular rules<br />Backward chaining<br />Certainty factors<br />RULE037<br />IF the organism<br />1) stains grampos<br />2) has coccus shape<br />3) grows in chains <br />THEN<br /> There is suggestive evidence (.7) that the identity of the organism is streptococcus.<br />
  22. 22. Mycin Consult<br />--------PATIENT-1--------<br />1) Patient's name: FRED SMITH<br />2) Sex: MALE<br />3) Age: 55<br />4) Have you been able to obtain positive cultures from a site at which Fred Smith has an infection? YES<br />--------INFECTION-1--------<br />5) What is the infection? PRIMARY-BACTEREMIA<br />6) Please give the date when signs of INFECTION-1 appeared. 5/5/75<br />The most recent positive culture associated with the primary-bacteremia will be referred to as:<br />--------CULTURE-1--------<br />7) From what site was the specimen for CULTURE-1 taken? BLOOD<br />8) Please give the date when this culture was obtained. 5/9/75<br />The first significant organism from this blood culture will be called:<br />--------ORGANISM-1--------<br />9) Enter the identity of ORGANISM-1. UNKNOWN<br />10) Is ORGANISM-1 a rod or coccus (etc.)? ROD<br />11) The gram stain of ORGANISM-1: GRAMNEG<br />. . .<br />Davis, et al., Artificial Intelligence 8: 15-45 (1977)<br />
  23. 23. How Mycin Works<br /><ul><li>Dynamically traces out behavior of (what might be) a flowchart
  24. 24. Information used everywhere appropriate
  25. 25. Single expression of any piece of knowledge
  26. 26. To find out a fact
  27. 27. If there are rules that can conclude it, try them
  28. 28. Ask the user
  29. 29. To “run” a rule
  30. 30. Try to find out if the facts in the premises are true
  31. 31. If they all are, then assert the conclusion(s), with a suitable certainty
  32. 32. Backward chaining from goal to given facts</li></li></ul><li>Explore Mycin’s Use of Knowledge<br />** Did you use RULE 163 to find out anything about ORGANISM-1?<br />RULE163 was tried in the context of ORGANISM-1, but it failed because it is not true that the patient has had a genito-urinary tract manipulative procedure (clause 3).<br />** Why didn't you consider streptococcus as a possibility?<br />The following rule could have been used to determine that the identity of ORGANISM-1 was streptococcus: RULE033<br />But clause 2 (“the morphology of the organism is coccus”) was already known to be false for ORGANISM-1, so the rule was never tried.<br />Davis, et al., Artificial Intelligence 8: 15-45 (1977)<br />
  33. 33. Even Simpler Representation<br />Disease<br />Disease<br />s1<br />s1<br />s2<br />s2<br />s3<br />s3<br />s4<br />s4<br />s5<br />s5<br />s6<br />s6<br />s7<br />s7<br />s8<br />s8<br />s9<br />s9<br />s10<br />s10<br />s...<br />s...<br />
  34. 34. Diagnosis by Card Selection<br />Disease<br />Disease<br />Disease<br />s1<br />s1<br />s1<br />s2<br />s2<br />s2<br />s3<br />s3<br />s3<br />s4<br />s4<br />s4<br />s5<br />s5<br />s5<br />s6<br />s6<br />s6<br />s7<br />s7<br />s7<br />s8<br />s8<br />s8<br />s9<br />s9<br />s9<br />s10<br />s10<br />s10<br />s...<br />s...<br />s...<br />Disease<br />s1<br />s2<br />s3<br />s4<br />s5<br />s6<br />s7<br />s8<br />s9<br />s10<br />s...<br />
  35. 35. Diagnosis by Edge-Punched Cards<br />Dx is intersection of sets of diseases that may cause all the observed symptoms<br />Difficulties:<br />Uncertainty<br />Multiple diseases<br />“Problem-Knowledge Coupler” of Weed<br />
  36. 36. Multi-Hypothesis Diagnosis<br />Set aside complementary hypotheses<br />… and manifestations predicted by them<br />Solve diagnostic problem among competitors<br />Eliminate confirmed hypotheses and manifestations explained by them<br />Repeat as long as there are coherent problems among the remaining data<br />
  37. 37. Internist/QMR<br /><ul><li>Knowledge Base:
  38. 38. 956 hypotheses
  39. 39. 4090 manifestations (about 75/hypothesis)
  40. 40. Evocation like P(H|M)
  41. 41. Frequency like P(M|H)
  42. 42. Importance of each M
  43. 43. Causal relations between H’s
  44. 44. Diagnostic Strategy:
  45. 45. Scoring function
  46. 46. Partitioning
  47. 47. Several questioning strategies</li></li></ul><li>QMR Database<br />
  48. 48. QMR Scoring<br />Positive Factors<br />Evoking strength of observed Manifestations<br />Scaled Frequency of causal links from confirmed Hypotheses<br />Negative Factors<br />Frequency of predicted but absent Manifestations<br />Importance of unexplained Manifestations<br />Various scaling parameters (roughly exponential)<br />
  49. 49. Example Case<br />
  50. 50. Initial Solution<br />
  51. 51. More Expert Systems<br />Causality?<br />What’s in a Link?<br />Temporal reasoning<br />Quantitative reasoning<br />Model-based reasoning<br />Workflow<br />
  52. 52. Meaning of Representation?<br />causes<br />S<br />D<br />Always? probability<br />Magnitude? severity; bad cold  worse fever?<br />Delay? temporality<br />Where? spatial dependency<br />Under what conditions? context<br />Interaction of multiple causes physical laws<br />Cross-terms high-dimensional descriptions<br />
  53. 53. Interpreting the Pastwith a Causal/Temporal Model<br />
  54. 54. Exploiting Temporal Relations<br />abdominal<br />pain<br />?<br />blood<br />transfusion<br />?<br />jaundice<br />transfusion precedes both abdominal pain and jaundice implies transfusion-borne acute hepatitis B<br />as in 1, but only by one day<br />jaundice occurred 20 years ago, transfusion and pain recent<br />Can be very efficient at filtering out nonsense hypotheses.<br />
  55. 55. The “Too Many Rules” Problem<br />One decision support rule may work very well.<br />10 rules may work better<br />1000 rules may be unworkable!<br />Need for a comprehensive “view” of the medical status of the patient<br />Alert based on a full set of diagnoses not individual facts<br />
  56. 56. Conclusions<br />Medical decision support can be helpful in reducing medical errors<br />Usually based on simple rule based systems<br />More advanced medical decision support systems can be use to help with diagnosis<br />Not much use of advanced systems at present in routine care<br />

×