Attenuation Coefficient for Hepatic Steatosis Using a Single Ultrasound System: Associations of Measurement Parameters With Interoperator Agreement and Diagnostic Performance

BACKGROUND. Clinical adoption of ultrasound attenuation coefficient (AC) measurements has been hindered by lack of a uniform measurement protocol and a range of factors that may cause variability. OBJECTIVE. The purpose of this study was to evaluate associations of ROI depth, ROI size, and confidence map threshold with interobserver agreement and diagnostic performance of ultrasound AC measurements in detecting and grading hepatic steatosis using MRI proton density fat fraction (PDFF) as the reference standard. METHODS. This prospective study enrolled adults with known steatosis or at risk for steatosis from October 2023 to August 2024. One of two operators obtained videos of AC acquisitions using a single ultrasound unit. Both operators independently reviewed all videos and placed circular ROIs to obtain AC measurements for all 24 possible combinations of four ROI depths (2.0, 2.5, 3.0, and 4.0 cm from liver capsule to ROI outer edge), three ROI sizes (3.0, 3.5, and 4.0 cm), and two confidence map thresholds (20% and 40%). Participants underwent MRI PDFF measurement as a reference. RESULTS. The analysis included 101 participants (mean age, 54.5 ± 12.1 [SD] years; 62 women, 39 men). Interoperator agreement was excellent for all combinations (intraclass correlation coefficient, 0.92-0.98). AC measurements showed strongest correlations (Spearman rank correlation coefficient, 0.81 and 0.80 for operators 1 and 2, respectively) with MRI PDFF at an ROI depth of 4.0 cm. The optimal combination considering correlations with MRI PDFF and AUC across steatosis grades included a depth of 4.0 cm, size of 4.0 cm, and threshold of 40%. This combination had an AUC for detecting steatosis with grade of greater than 0, greater than 1, and greater than 2 for operator 1 of 0.93, 0.88, and 0.81, respectively, and for operator 2 of 0.92, 0.86, and 0.81, respectively. However, accuracy for detecting steatosis (grade > 0) was highest for the combination of depth of 3.0 cm, size of 4.0 cm, and threshold of 40% (operator 1, 90.1%; operator 2, 82.2%). CONCLUSION. AC measurements showed excellent interoperator agreement across parameter combinations. Correlations with MRI PDFF were strongest at a depth of 4.0 cm. Combinations yielding highest diagnostic performance were identified. CLINICAL IMPACT. These results will help determine a standardized optimal protocol for ultrasound AC measurements, facilitating clinical adoption for liver fat quantification.