Probability density in the analysis of medical images

Document Type

Conference Proceeding

Date of Original Version



An extension to a probability density function (PDF)-based analysis of medical images is discussed. The PDF of interest is that of fractional Brownian motion, and the extension is a reformulation of the problem to explicitly exploit two-dimensional data. It is found that the resulting segmentation can be much more accurate than previous methods. However, due to the computational intensity of the problem, the algorithm results in a binary segmentation, rather than the 1:1 mapping, from fractal dimension to gray level, that was intended. Regardless of the particular type of segmentation produced, the sensitivity of the segmentation in the test cases shows great promises for this approach. In the present application, the images are magnetic resonance images of human hearts.

Publication Title

Proceedings of the IEEE Annual Northeast Bioengineering Conference, NEBEC