A Hybrid Computational Intelligence Approach Combining Genetic Programming And Heuristic Classification for Pap-Smear Diagnosis

Athanasios Tsakonas, Georgios Dounias, Jan Jantzen, Beth Bjerregaard

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

The paper suggests the combined use of different computational intelligence (CI) techniques in a hybrid scheme, as an effective approach to medical diagnosis. Getting to know the advantages and disadvantages of each computational intelligence technique in the recent years, the time has come for proposing successful combinations of CI tools and techniques for the improvement of decision making, Diagnosis and classification in complex domains of application. In the current approach genetic programming is embedded within a heuristic scheme for classification of medical records into different diagnoses. The final result is a short but robust rule based classification scheme, achieving high degree of classification accuracy (exceeding 90% of accuracy for most classes) in a meaningful and user-friendly representation form for the medical expert. The domain of application analyzed through the paper is the well-known Pap-Test problem, corresponding to a numerical database, which consists of 450 medical records, 25 diagnostic attributes and 5 different diagnostic classes. Experimental data are divided in two equal parts for the training and testing phase, and 8 mutually dependent rules for diagnosis are generated. Medical experts comment on the nature, the meaning and the usability of the acquired results.
Original languageEnglish
Title of host publicationProc. European Symposium on Intelligent Technologies, Hybrid Systems and their implementation on Smart Adaptive Systems, EUNITE
Number of pages10
PublisherELITE Foundation, Pascalstrasse 69, D-52076 Aachen, Germany
Publication date2001
Publication statusPublished - 2001

Cite this