Comparison of human observer efficiency in pelvis radiographs in two different anatomical regions
2004 (English)Report (Other academic)
The performance of an x-ray imaging system is often evaluated in terms of how well low-contrast details are visualised in the images. It is however difficult to obtain reliable results because a threshold for human visibility does not strictly exist. The degree of how well a low-contrast detail is ‘seen’ varies from ‘not seen’ to ‘clearly seen’ related to the certainty (or level of confidence) of detection. The visibility criterion is difficult to define, communicate and maintain which results in comparably large inter- and intraobserver variability.
The detection probability depends not only on the imaging task, the technical performance of the imaging system and the observer´s skill and training but also on the projected anatomy in the region where the detail (i.e. lung nodule) is situated. This has been explored by Kundel et al (1985), Samei et al (1999), Håkansson et al (2004), and in Båth et al (2004) by ROC studies mainly in chest radiography.
In this work, human observer detection efficiency was measured in a pelvis anthropomorphic phantom. Low-contrast lesions were constructed and positioned in two different regions, one region with a fairly homogeneous, the other region with a heterogeneous anatomical background. A group of human observers were asked to identify the lesion in a set of two-alternative forced choice (2-AFC) experiments.
Place, publisher, year, edition, pages
Linköping: Linköping University Electronic Press , 2004. , 6 p.
Report / Department of Radiology, Linköping University, 1998-, 101
Radiology, Nuclear Medicine and Medical Imaging
IdentifiersURN: urn:nbn:se:liu:diva-57877ISRN: LIU-RAD-R-101OAI: oai:DiVA.org:liu-57877DiVA: diva2:328215