Skip to Main Content
Cardiac Echo-Doppler (Echo) data may contain hidden information that cannot be revealed and identified by an experienced cardiologist. Thus, important relations between cardiac dimensions (CD) may be misinterpreted. Clustering is commonly used in Data Mining (DM) and aimed to partition data into clusters. The aim of this work was to find possible correlations between CD in order to upgrade and improve diagnostic abilities. This relation was already proved for homogenous populations (and known mechanism) of diseases such as Hypertension. In contrast, in our work the population was heterogeneous (normal and various diseases) with no known mechanism (such as Hypertension). Therefore, clustering algorithms such as K-means (KM), Kohonen (Koh) and TwoStep (TS) were applied on 24,400 data objects of Cardiac Echo measurements. The commercial DM tool Clementine (Clem) was used. Each algorithm generated different clusters. Despite this, between left atrial area (LA_Area) and ascending aortic diameter (AsAo_Dia), pathological correlations were identified.