Abstract
A consistent, quantitative, observer-independent method of characterising mammographic parenchymal pattern is described. The method is based on the calculation of the 'fractal dimension' of digitised mammograms. The degree of correlation between the parenchymal pattern classifications by a fractal-based system and those of radiologists is assessed. For a set of 70 mammograms, average weighted proportion agreement among three radiologists in calling Wolfe grades was 85%, while agreement between the radiologists and the fractal classifier was 84%. The method developed may prove to be useful in establishing an index of risk for breast cancer and, ultimately, in determining intervals between examinations for individuals in a mammographic screening programme.