Computerized Decision Support for Medical Imaging

1
Computerized Decision Support for Medical Imaging

• Authors C. Shyu, C.Brodley, A. Kak, A. Kosaka,
  A. Aisen, L. Broderick
• Journal Computer Vision and Image Understanding
  , Vol. 75, 1999

2
Computerized decision support systems

• Active knowledge systems which use two or more
  items of patient data to generate case-specific
  advice.
• knowledge systems
• Knowledge source from which the advice is picked
  images, algorithms for manipulating the images,
  medical data, derived probabilities and symbolic
  representation of medical facts
• A mechanism by which a user may quickly derive
  relevant information from the knowledge source

3
Computerized decision support systems

• General criteria four successful decision support
  systems
• Need for a decision aid
• Ex when the interpretation requires specialist
  expertise, when many images are generated or
  interpretation is specially difficult because of
  noise or because the image is visually
  complicated
• Practicality
• the constraints of the medical domain and the
  clinical setting should be taken into account

4
Computerized decision support systems

• General criteria four successful decision support
  systems (cont.)
• Veracity
• The knowledge source on which the decision
  support is based must be accurate and complete
• Relevance
• The system provides the user with information
  which improves his/her decision making.

5
Computer aids for diagnostic radiology

• Image databases
• Decision support systems based on numerical
  methods
• Use of Bayes rule to combine, for example,
  information about the frequency of a disease and
  the frequency with which signs are associated
  with that disease
• Expert Systems
• they are based on a set of rules which represent
  the knowledge used by clinicians in making
  decisions. The rules are used together with
  information provided by the user to generate
  inferences.

6
Architecture of a medical imaging expert system
7
Applications of the Imaging Expert Systems

• Automatic generation of anatomical and functional
  atlases
• CBIR
• The Imaging expert systems should improve the
  diagnostic accuracy by providing a second
  opinion they should provide objective measures
  or normal and abnormal patterns and dray the
  attention of radiologists

8
Computer Vision in Medicine

• Automatic detection of clustered
  micro-calcifications on mammograms
• Computerized detection of pulmonary nodules
• Computerized analysis of heart size
• Characterization of interstitial disease on chest
  radiographs
• Automatic tracking of vessels on angiographic
  images

9
Computer Vision in Medicine

• Low-level features were extracted for the
  different applications, but they were limited to
  a certain disease, or organ or pathology
• Therefore, the low-level feature extractors
  should be combined and included in future imaging
  expert systems.
• Multiplying the inputs increases the quantity of
  information the expert system can use to support
  diagnosis hypothesis.
• Spatial relationships between features

10
Future trends in Medical Imaging

• A new framework for a natural and systematic way
  of gathering knowledge from domain experts.
• Integration of the IES into the PACS systems.
• Some preliminary results for digital mammography.
• New standards to measure the accuracy of the IES
• Subjective assessment of radiologists is often
  the only reference available

11
Future trends in Medical Imaging

• Future applications from pure diagnostic purpose
  to therapeutic management
• Taking into account the reaction of the patient
  to a certain therapy will provide information
  for supporting or refining the diagnostic
  hypothesis.
• T-HELPER
• Medical training
• Quick medical reference ---electronic textbook of
  medicine
• INTERNIST-1

12
Computerized decision support in diagnostic
radiology

• What is an expert system?
• A computer system that is programmed to imitate
  the problem-solving procedures that a human
  expert makes. For example, in a medical system
  the user might enter data like the patient's
  symptoms, lab reports, etc., and derive from the
  computer a possible diagnosis. The success of an
  expert system depends on the quality of the data
  provided to the computer, and the rules the
  computer has been programmed with for making
  deductions from that data. (High-Tech
  Dictionary)