1 / 13

A new medical test for for atherosclerosis detection GeNo

ECML/PKDD-2003. A new medical test for for atherosclerosis detection GeNo. Jérôme Azé, Noël Lucas, Michèle Sebag (LRI, Orsay). Approach. Goal : PREVENTION OF CV DISEASES  IDENTIFICATION OF RISK FACTORS 

sidone
Télécharger la présentation

A new medical test for for atherosclerosis detection GeNo

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ECML/PKDD-2003 A new medical test for for atherosclerosis detection GeNo Jérôme Azé, Noël Lucas, Michèle Sebag (LRI, Orsay)

  2. Approach • Goal : PREVENTION OF CV DISEASES  IDENTIFICATION OF RISK FACTORS  BUILD A SENSITIVIE INEXPENSIVE MEDICAL TEST • Questions : - WHICH FACTORS ? - HOW DO THEY INTERACT ?

  3. OUR GOAL • Usual ML & KDD • Provides expert with hypotheses : • EXPERT says YES or NO. • Our approach (Visual Data Mining) • Provides the expert with curves : • EXPERT says « let me think… » USE VISION TO THINK ( Card, Mackinlay, Schneiderman, 1999 )

  4. OUR GOAL, cont’d 1/ We need a fine-grained perception of risk Classification  Regression 2/ Only sufficient (not necessary) conditions can be extracted from data bases. Ex. : disease if In DB : your parents are diabetic In DB : You are smoking for long You don’t laugh often enough

  5. Contents • Motivations • Univariate and Multivariate Analysis • Presentation of ROGER : ROc based GEnetic learneR • Validation and construction of a Medical test • Perspectives

  6. Univariate Analysis • graphical analysis of one risk factor for the healthy and ill patient (of the risk group) • limitations : factors interact

  7. Multivariate Analysis • limitations : manual selection of factors

  8. Presentation of ROGERROc based GEnetic learneR Search space / hypothesis space linear function of attributes h(Ex) = i x atti (Ex) avec (Ex, +/-) Evaluation of an hypothesis : area under the ROC curve h  (risk(Ex), Etiq(Ex)) Sort patient by increasing risk ++++-+---++-----+++----------- + : ill - : healthy

  9. The ROC curve(Receiver Operating Characteristics) Evaluation of a medical test : • True positive rate (medical sensitivity) • True negative rate (medical specificity) True Positive rate False positive rate

  10. Validation and construction of a Medical test • goal: an inexpensive test (exploiting available information) Test 1: the whole set of attributes Test 2: no medical attributes Test 3: medical attributes + bmi, age

  11. validation of the first test • target : patient of the normal group, with maximal risk estimate • study of the patients with risk > risk of the target

  12. Conclusions • from the learning and mining view point • relevant approach for feature selection and sensitivity analysis • from a medical perspective • satisfactory performance • Demonstrate the contribution of habitus to CV disease

  13. Perspectives • from the learning and mining view point • the variability of evolutionary solution can be use in a visual data mining framework • from a medical view point • hypotheses suggested by learning and mining must be validated within medical procedure

More Related