Docstoc

R4

Document Sample
R4 Powered By Docstoc
					      R4

Andy Aiken
Overview
 The Refiner series
 Cognitive psychology of diagnosis
 Diagnosis Web
 Proposed work: R4
The Refiner Series
   Case-based system
   Cases are assigned to categories
   System generates descriptions of these
    categories
   Resolves inconsistencies in datasets
     An inconsistency is when a case does not only match
      the category it has been assigned to by an expert
     User-directed
The Refiner Series
   History
     Refiner:initial system (1988)
     Refiner+: batch system (1995)
          Much more efficient
     Refiner++:    current version
        Java, faster, more strategies and data types, GUI
        Evolution not revolution
The Refiner Series
The Refiner Series
   Refiner++ was shown to experts
     Anaesthetics,child psychology, HDU, medical
      CAL, engineering
     Reaction was generally positive
         System to be used in several proposed
          interdisciplinary projects
     BUT…
The Refiner Series
   Limitations
     Time    Series
          Many medical datasets, for example, are time-
           series data
     Noise
        Can deal with noise, but in a limited and
         sometimes confusing way
        This is our focus
Refiner and Noise
 What changes are needed?
 Changes the focus of the system
     Refiner++:  the creation of category
      descriptions
     R4: classification

   This requires us to investigate the science
    of diagnosis
Diagnosis
   How do clinicians perform diagnosis?
     How is it taught?
     What is the cognitive psychology?
Diagnosis
   Differential Diagnosis
     What  is a differential diagnosis?
     Requires the diagnostician to memorise a
      vast amount of information
     Simplistic view of the process
Diagnosis
Diagnosis
   Hypothetico-Deductive Reasoning
     Uses   both forward and backward chaining
     Iterative
     Requires significant cognitive processing
Diagnosis
   Given initial facts,
    work forward to likely
    hypotheses
   These are the
    differential diagnoses   Facts   Hypotheses
Diagnosis
   Identify other facts which
    should be present for
    these hypotheses to hold,
    and investigate them
   Can attempt to
    distinguish between a set    Facts   Hypotheses

    of competing hypotheses,
    or get additional
    confirmation for a
    hypothesis
Diagnosis
   Repeat until satisfied
     One  hypothesis is
      notably more likely
      than others
     One likely hypothesis
                                  Facts   Hypotheses
      is especially significant
     The courses of action
      are similar
Diagnosis
   The effect of expertise
     Novices  perform HDR, but the process is
      slow, potentially incomplete and inaccurate
     Intermediates perform HDR, but this is still a
      slow process
     Experts still perform HDR but this is
      supplemented by compiled knowledge
      (shortcuts; faster)
Diagnosis
   Illness Scripts (Feltovich, 1984)
     Compiledknowledge used in diagnosis
     Comprises three knowledge sources:
        Faults
        Consequences

        Enabling conditions

     Says   nothing about how they are used
Diagnosis
   Faults
     Theunderlying malfunctions which give rise to the
      symptoms
   Consequences
     Detectable   signs and symptoms
   Enabling conditions
     Background knowledge
     Predisposing factors (patient’s case history)
     Boundary conditions (age and gender)
     Hereditary   factors
Diagnosis
 Illness scripts don’t explain the process of
  diagnosis
 Diagnosis Web
    A  more inclusive explanation
     More knowledge sources
     Links between them indicate processes
Diagnosis Web
                                                 Courses of Action



                         Hypotheses




   Cases                 Contextual                   Faults
                          Factors




           Descriptors                Features
Diagnosis Web
   How do we use the diagnosis web to
    improve the Refiner series?
R4
   Will be aimed at novice diagnosticians
     Medical  students, but not limited to medical
      domains
     An aid to diagnosis
     Supports the student’s reasoning without
      taking over
R4
   Three stories:
     An aid to diagnosis
     Scaffolding user’s diagnostic skill
     Generation of test cases
R4
 User presents a case to the system
 The system displays two lists:
     Likely  (differential) diagnoses
     Potential next steps (descriptors to find values
      for) to hone in on the correct diagnosis
   Providing more data for the case updates
    these lists automatically
DXplain
DXplain
 Used by clinicians to perform diagnosis
 User enters findings, system updates lists
  of likely diagnoses and other findings to be
  investigated
 Sound familiar?
R4 and DXplain (et al)
   Superficially very similar
   R4 is case-based; others are database-driven
     Easier   to create & maintain a casebase
   R4 uses descriptor values; others are term-
    based
     Lends    itself to automation
   No other system attempts to mirror the cognitive
    process of diagnosis
Conclusion
   Refiner++ was well received but did not handle
    noisy datasets well
   Adding the facility to handle noisy datasets
    changes the focus to classification
   Combining HDR and illness scripts leads to the
    Diagnosis Web
   The new Refiner system will incorporate aspects
    of HDR and illness scripts to mirror how
    diagnosis is performed
Questions?

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:6
posted:9/2/2012
language:English
pages:29