Clustering of time activity curves for uptake pattern assessment in dynamic nuclear medicine imaging

Abstract

Nuclear medicine instrumentation visualize the radiopharmaceutical uptake inside the body allowing the interpretation of physiological processes. In dynamic nuclear medicine imaging, time-dependent image sequences are recorded. The changes of radiopharmaceutical uptake over time (so calles time activity curves, TACs) can be analyzed in order to find abnormal patterns corresponding to either structural or functional disorders. Hierarchical Cluster Analysis (HCA) is a powerful statistical tool for classification. We applied HCA on TACs to find clusters of similar TAC patterns. Optimal number of clusters is determined by Hubert's rule. We used Principal Component Analysis (PCA) on TAC clusters to find a representative TAC that presents the uptake pattern in the region of each cluster. The application of algorithm is illustrated in the patient with the histopatologically proven parathyroid hyperplasia, but the developed tool is useful for finding the appropriate classification method of TAC patterns in all types of dynamic nuclear medicine studies.