Abstract
The purpose of the present study was to evaluate the relevance of using a phantom to simulate a clinical situation where small low contrast objects are embedded in relatively homogeneous organs in order to discriminate between different ultrasound machines, taking into account human observer variability. One high-end and one general ultrasound machine using the same probe were included. Images containing 4 and 6.4-mm objects of four different contrasts were collected from a greyscale phantom at different depths. Six observers participated in a 4-alternative forced choice study based on 960 images. Variability was determined using bootstrapping. At four of sixteen depth/size/contrast combinations, the visual performance of the high-end machine was significantly higher. The results indicate that it is possible to use a greyscale phantom to discriminate between ultrasound machines in terms of their ability to reproduce clinically relevant low-contrast objects. However, the number of images and number of observers needed are larger than those usually used for constancy control, if the large uncertainties caused by human observer variability is to be taken correctly into account.