A simulation model of clustered breast microcalcifications.

A simulation model of clustered microcalcifications superimposed on normal mammographic backgrounds has been developed and evaluated. A cluster is described according to its size, the number of microcalcifications it contains, and their density. A microcalcification is defined on the basis of its size, shape, location within the cluster, its contrast, and the distribution of the gray-level values between pixels that belong to it. Gray-level parameters are dependent on the surrounding tissue, which is also slightly modified. The main dependencies between the parameters are taken into account. All the parameters are randomly sampled using distribution laws determined from the statistical analysis of 408 real clusters containing a total of 8611 microcalcifications. Some statistical characteristics depend on the structural mammographic pattern on which the cluster is superimposed. The evaluation of the simulation model, performed by two radiologists on a test set of 100 real and simulated clusters, did not reveal any statistically significant difference between the simulated clusters and the real ones. This method provides clusters of microcalcifications with well-defined characteristics and can be used to evaluate the accuracy of computerized detection methods.