Evaluation of subjective and objective ultrasound quality assurance sensitivity assessment methods
Tuathan O’Shea, Joint Dept. of Physics, The Royal Marsden NHS Foundation Trust, London
Purpose:
Current quality assurance (QA) procedures incorporate subjective tests of system sensitivity, i.e., the ability of the system to detect and display weak echoes. While it is apparent that automated methods could reduce the time intensiveness of QA, it is not clear to what extent automation influences the variance and also which sensitivity method is most reproducible.
Method:
Two routine system sensitivity tests were evaluated namely the in-air reverberation distance (AR)(IPEM 102 guidelines) and the depth of penetration (DOP)(AAPM TG1 guidelines) for linear array probes from three ultrasound scanners (SuperSonic Imagine Aixplorer/SL15-4(S1), Medison Accuvix XQ/L5-12IM(S2) and Zonare Z.one/L10-5(S3)). Following guidelines, four observers manually performed (×3) the tests and saved images for automated analysis. A computer program was developed to automatically estimate the AR and DOP using observer images. Results were compared statistically (T-test) and by coefficient of variation (COV).
Results:
Mean inter-user manual and automated AR and DOP distances were either not statistically different (p>0.05)(S2/S3) or within the tolerance of one reverberation line (AR,~1mm) or 1cm (DOP) presented in guidelines (S1). The mean COV difference between manual and automatic methods was also not significant. The mean COV was lower for both manual and automated estimates of AR (5.3% and 4.0%, respectively) compared to DOP (8.5% and 8.7%). For a single observer and imaging preset, mean (intra-user) COV was <5% for both manual and automated estimates of AR and DOP. AR was more sensitive than DOP to -3dB reduction in transmit power (mean change in AR was -8.6±6.2% greater).
Conclusions:
Results support the use of automated and objective estimates of sensitivity. Automatic methods would speed-up QA. Automatically reproduced image acquisition settings such as a QA preset could also be employed, due to removal of subjectivity in image interpretation. AR estimates appear to be more sensitive to changes in scanner sensitivity than DOP.