A computerised quality control testing system for B-mode ultrasound

Current methods of ultrasound (US) imaging quality control involve an observer taking a series of measurements on test object images. This process is very labour-intensive, rapidly becoming more so as the complexity of US scanners increases. Also, many of the measurements are based on the subjective judgement of the operator and are, therefore, prone to an undesirable level of intraobserver and interobserver variation. We have developed a suite of programs to analyse captured US images to estimate a number of performance parameters, including resolution, high- and low-contrast penetration depths and high- and low-contrast sensitivity. In this system, images of the ATS Laboratories Multipurpose Phantom Model 539 are acquired on a PC with a video capture card, and analysed using the software developed. Automated tests have been developed for resolution, low- and high-contrast penetration and low- and high-contrast sensitivity. A preliminary validation of the system was carried out using images from two curvilinear probes and a linear-array probe operating at 5 MHz, 7.5 MHz and 10 MHz, respectively. The 95% confidence intervals ranged from ± 17% to ± 21% for resolution measurements and from 0% to ± 1.5% for low-contrast penetration. The system provides an index of visibility for high- and low-contrast targets, as compared to the coarse visual assessment of a human observer, with 95% confidence intervals ranging from ± 6% to ± 13% for low-contrast sensitivity and from ± 3% to ± 25% for high-contrast sensitivity. Advantages of the system over the human observer include improved reproducibility and increased information regarding visibility of targets. A more detailed multicentre validation is currently being undertaken. (E-mail: ngibson@ncht.org.uk)