Introduction of weighted structural complexity measures for analysis of noisy signals and introduction to nuclear medicine image processing

In this work a sensitive source detection algorithm for nuclear medicine imaging is presented. Special attention is given to weak sources with strength comparable to fluctuations of the underlying noise. The scaling-index-method (SIM) originally developed by Morfill et al. (Max-Planck-Institute for Extraterrestrial Physics) has been extended. The SIM assumes that the local point density in a sphere with radius r around a point in a discrete phase space follows the scaling law rα. For a digitized image F(x,y) the points of this space are represented by the coordinates (x,y) and the gray-level F(x,y). Because the scaling behavior of noise and signal points differs, discrimination between sources and noise is possible. In this work the author extends the SIM by applying weighting functions to the locally calculated scaling indices, which take into account the information available on the noise characteristics. Because of the poor photon density in nuclear medicine diagnostics the image quality is highly influenced by the statistical fluctuations inherent in radioactive decay. Consequently, only those changes in radiopharmaceutical distributions will be detectable for the observer, which are greater than statistical noise. On the other hand the detection of weak sources is one of the most important tasks in nuclear medicine diagnostics. Therefore in several clinical cases, such as total-bone or thyroidal scintigraphy, the applicability of the developed algorithms is studied. This work documents that by using the proposed methods even weak sources hardly detectable for the examining physician are identified with a high significance. Furthermore the SIM allows to discriminate real sources from tomographic artefacts