A mathematical method for constraint-based cluster analysis towards optimized constrictive diameter smoothing of saphenous vein grafts.

This study was concerned with the cluster analysis of saphenous vein graft data to determine a minimum number of diameters, and their values, for the constrictive smoothing of diameter irregularities of a cohort of veins. Mathematical algorithms were developed for data selection, transformation and clustering. Constrictive diameter values were identified with interactive pattern evaluation and subsequently facilitated in decision-tree algorithms for the data clustering. The novel method proved feasible for the analysis of data of 118 veins grafts, identifying the minimum of two diameter classes. The results were compared to outcome of a statistical recursive partitioning analysis of the data set. The method can easily be implemented in computer-based intelligent systems for the analysis of larger data sets using the diameter classes identified as initial cluster structure.