A Pre-Release Algorithm With a Confidence Map for Estimating the Attenuation Coefficient for Liver Fat Quantification.

ABSTRACT

OBJECTIVES: To compare the estimates of attenuation coefficient (AC) for liver fat quantification between 2 Ultrasound systems and to evaluate the quality measure of a pre-released software. METHODS: AC were obtained in 30 participants in this single-center IRB-approved, HIPAA compliant study. Images were obtained on the Philips Epiq Elite system using experimental software and the Canon Medical Systems Aplio i800 with released software. Five AC measurements were taken and the median and IQR/M were calculated. Region of interest placement was based on a confidence map. ROI was at the same depth and size for each system. The concordance was estimated using the Lin's concordance correlation coefficient (CCC), the r Pearson's correlation coefficient, the bias-correction factor (Cb), and the Bland-Altman method. RESULTS: The ACs varied from 0.45 to 1.0 dB/cm/MHz for the Philips system and 0.30 to 0.96 dB/cm/MHz for the Canon system. The CCC (95% CI) was 0.792 (0.666-0.918), Pearson's r was 0.839 with Cb of 0.944, and the mean difference was 0.03 (-0.101; 0.162) suggesting the 2 methods are considered to be in agreement. Based on a Philips confidence map to determine the best location for performing the measurements, a depth of 3.5 to 4.0 cm from the liver capsule was determined, which might be significantly different than that of the Canon system. CONCLUSIONS: Estimation of the AC of the 2 systems showed a high agreement, that is, a similar trend. Assessment of the placement of the measurement box based on the quality of the measurement might be different between the 2 systems.