Skip to Main Content
Developing models for microcalcification clusters remains difficult because of the variety of shapes and orientations, background tissue structures and image qualities. A model which has been applied successfully to medical images is fractional Brownian motion. The fractal dimension, scale-invariance and self-similarity properties of fractional Brownian motion closely resemble many natural phenomena such as growth patterns. These properties are investigated and several fractal dimension estimation algorithms are presented for both 1-D and 2-D signals. A selection of these estimators is applied to the problem of malignant and benign microcalcification cluster discrimination. Results indicate that modelling mammographic images with fractional Brownian motion can provide superior indicators of the global texture than measures based on traditional texture features.