What is it about?
In the process of diagnostic classification, enormous quantities of numerical details are processed and considered at different levels of data abstraction, and particular data quantification and different data aggregation are performed. In order to understand this process, it is important to model and formalize it with expressive schema/tools, for an accurate approximation of the real process and for an ‘efficient learning.’ In this paper, first we describe the organization and the characteristics of the ‘information granular computing’ schema, describing the main components of it, with particular emphasis on descriptive models and predictive models. First, we present a real complex problem of diagnostic classification: the computerized ECG classification examined with two validated databases. Then, three descriptive models will be described in detail: self-organized maps (SOM) model, the radial basis Functions (RBF) model, and the linguistic model. For every model, the application to ECG diagnostic classification will be considered, and the corresponding accuracy of the entire classification task will be described and compared.
Featured Image
Why is it important?
The different abstraction levels of the process of diagnostic classification in medical field are considered in the framework of granular computing. In this process, three different and characteristic blocks can be individuated: (1) quantification phase, (2) descriptive models, and (3) predictive models. Three descriptive models will be described in detail: self-organized maps (SOM) model, the radial basis Functions (RBF) model, and the linguistic model.
Perspectives
In this chapter, different levels of data abstraction in the process of diagnostic classification in medical field have been studied. These levels were analyzed and discussed in the framework of information granular computing, and the following two main blocks were described in detail: the descriptive model and the predictive model. Three representative examples of information granulation have been considered: the SOMmodel, theRBFmodel, and the linguistic model.Areal complex problem of diagnostic classification was considered for testing and validating the proposed approaches: the computerized ECG classification with two validated databases. The description, the analysis, and the discussion of the three methods have shown that the individualization of different abstraction levels and the use of information granule in different way have improved the generalization capability, the accuracy, and the transparency of the classification task.
Dr Giovanni Bortolan
IT CNR
Read the Original
This page is a summary of: Granular Computing in Medical Informatics, Wiley,
DOI: 10.1002/9780470724163.ch39.
You can read the full text:
Contributors
The following have contributed to this page
