Fuzzy Clustering: A Versatile Mean To Explore Medical Databases

G. Berks, Diedrich Graf von Keyserlingk, Jan Jantzen, M. Dotoli, H. Axer

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearch

    Abstract

    A symptom is a condition indicating the presence of a disease, especially, when regarded as an aid in diagnosis.Symptoms are the smallest units indicating the existence of a disease. A syndrome on the other hand is an aggregate, set or cluster of concurrent symptoms which together indicate the presence and the nature of the illness. To each syndrome belongs a history with the first description, confirmation, and acknowledgement of its usefulness and in many cases dedicating of its name to the first author. Looking for concurrent symptoms is now as ever the main task of diagnosis. Classification and clustering are the basic concerns in medicine. Classification depends on definitions of the classes and their required degree of participant of the elements in the cases' symptoms. In medicine imprecise conditions are the rule and therefore fuzzy methods are much more suitable than crisp ones. Fuzzy c-mean clustering is an easy and well improved tool, which has been applied in many medical fields. We used c-mean fuzzy clustering after feature extraction from an aphasia database. Factor analysis was applied on a correlation matrix of 26 symptoms of language disorders and led to five factors. The factors displayed meaningful indication of the disease. After the factors are gained there are usually treated with the so-called varimax method and transformed into 'simple structure' to render easier interpretation of their significance. This additional work allows to ensure the statistical validity of the factors but methods have a disadvantage of being empirical in nature. It is obvious that the information contained in the factors are already present in the original state, that is before transformation. The extracted factors are polarized. The loading of the factors may be transferred to membership functions of symptoms. The symptoms reveal in this way different membership to different aspects of language disorders. Fuzzy c-mean clustering was used to advise the symptoms to the different categories, because of polarization of the five factors at least 10 categories result from this proceeding. A symptom may belong to more than one class. For instance to the class of very severe disease and to the class of failure of awareness of the own disturbance. The description of language failures by c-mean classification of analyzed factors correspond in many but not in all cases to the traditional diagnostic scheme.
    Original languageEnglish
    Title of host publicationFuzzy Clustering
    Publication date2000
    Publication statusPublished - 2000
    EventEuropean Symposium on Intelligent Techniques - Aachen, Germany
    Duration: 14 Sep 200015 Sep 2000

    Conference

    ConferenceEuropean Symposium on Intelligent Techniques
    CountryGermany
    CityAachen
    Period14/09/200015/09/2000

    Cite this