Supporting an Early Detection of Diabetic Neuropathy by Visual Analytics

Abstract

In this paper, we describe a step-wise approach to utilize ophthalmic markers for detecting early diabetic neuropathy (DN), the most common long-term complication of diabetes mellitus. Our approach is based on the Visual Analytics Mantra: First, we statistically analyze the data to identify those variables that separate DN patients from a control group. Afterwards, we show the important separating variables individually, but also in the context of all variables regarding a pre-defined classification. By doing so, we support the understanding of the categorization in respect of the value distribution of variables. This allows for zooming, filtering and further analysis like deleting non-relevant variables that do not contribute to the definition of markers as well as deleting data records with false data values or false classifications. Finally, outliers are observed and investigated in detail. So, a third group of potential DN patients can be introduced. In this way, the detection of early DN can be effectively supported.