Abstract: | The authors describe a method of treating scintigraphic data with which they can correctly compare images of a structure that were obtained in different conditions. This method comprises two steps which are fully computerized. During the first one known as the "registration" step and intended to make a comparison possible they maximise a new similarity criterion by a series of random trials in a five parameter space. During the second, which is the actual comparison, they use an appropriate statistical test to recognise those homologous pixels with a significantly different contents which will be the only ones retained to build up the final image. They give then two examples of what the method brings in the medical field which are the detection of adenomatous parathyroid glands and the follow-up of lung perfusion in case of embolism. |