Synthesis of texture from clinical images

Psychophysical tests are a major tool in medical imaging for accessing a persons performance in viewing medical images. Tests that use clinical images are preferred as the results are more representative of the clinical task but these require large numbers of images and accurate definition of the gold standard. In this paper methods for synthesising clinical textures were evaluated. One non-parametric method was found to be more suitable although clinical samples were difficult to synthesise. A modification, based on sample reordering was developed. Results are presented showing the success of this approach for a range of clinical images and modalities. This may help facilitate psychophysical experiments where anatomical background noise is to be considered. (C) 2003 Elsevier Science B.V. All rights reserved.