Evaluating the Performance of Association Mining Methods in 3-D Medical Image Databases
Vasileios Megalooikonomou · 2002
Evaluation of the process of mining associations is an important problem in database systems and especially those that store critical data and are used for making critical decisions. In the context of spatial databases and in particular in 3-D medical image databases we present an evaluation framework in which we use probability distributions to model the spatial regions of interest (ROIs), and Bayesian networks to model the joint probability distribution among ROIs and observed deficits or medical conditions. By controlling these parameters, we evaluate as example, the Fisher exact test of independence, one of the methods currently available for detection of associations. We obtain measures of recovery of known associations as a function of the number of samples used, the strength and number of associations in the statistical model, the number of spatial ROIs associated with a particular deficit, the prior probabilities of spatial regions being of interest, the conditional probabilities of the deficits and the spatial normalization error.