A CDMAM image phantom software improvement for human observer assessment

Abstract

A software tool is presented to improve the features of CDMAM image phantom by University Hospital Nijmegen. This software tool ensures that the 4-alternative forced choice method of CDMAM is actually kept, even when is being scored by highly expertise observers familiar on the test object pattern. For digital images, the developed software tool automatically changes the image position of the four corners. It can be selected a fixed rotation angle or a random one, so making impossible that any observer is able to remember the exact corner position of the target disc inside any cell. Two alternative successful algorithms have been tested. ROC curve analysis obtained by 36 observers shows that both original and computer-modified images are indistinguishable. The ROC area was 0.507±0.024 for first algorithm and 0.522±0.026 for the second one, indicating that there was no statistical difference between real and computer-modified images for both of them. © 2008 Springer-Verlag Berlin Heidelberg.