Itemset tree

For a pattern analysis with frequent itemset mining, the observed plaque distribution has to be described meaningfully by items. The whole set of items created per patient is referred to as an observation. For the plaque position, a multi-scale approach in form of an itemset tree was developed to assign several items per plaque position. This allows the statistical analysis within and between different levels, i.e. it is possible to extract correlations between various segments but also between segments and certain areas (left, right, proximal, distal,...) of the coronary tree. Based on the 16-segment coronary artery scheme of the American Heart Association, we developed an itemset tree that consits of six hierarchical levels (see paper).


16-segment coronary artery scheme as proposed by the American Heart Association.