A reproducible ultrasound localization microscopy framework for the quantitative imaging of hepatocellular carcinoma on a clinical ultrasound system.
Poor differentiation and microvascular invasion (MVI) are crucial prognostic indicators for hepatocellular carcinoma (HCC), but evaluating them before surgery is still difficult. Our goal was to develop a reproducible framework for ultrasound localization microscopy (ULM) using a clinical ultrasound system and assess its effectiveness in predicting tumor differentiation and MVI.
Participants were prospectively enrolled and underwent contrast-enhanced ultrasound and ULM imaging. We first compared frame selection strategies by calculating coefficients of variation (COVs) of ULM parameters derived from motion-variance curve (MVC) alone versus MVC combined with time-intensity curve (TIC). ULM resolution and parameter stability were then assessed across low (0-399), medium (400-699), and high (700-1000) frame counts. Regions of interest (ROIs) were manually drawn on grayscale ultrasound and mapped to ULM images. Inter-operator agreement was evaluated using intraclass correlation coefficients (ICCs). Participants were grouped by pathological differentiation and MVI status. The predictive performance of ULM parameters was assessed with multivariable logistic regression.
Sixty-one HCC participants were enrolled (11 poorly differentiated, 50 well differentiated; 30 MVI-positive, 31 MVI-negative). The MVC+TIC strategy yielded significantly lower COVs, indicating higher repeatability of ULM parameters. Medium frame counts (400-699) provided high resolution (minimum vessel diameter 91.3 ± 22.7 μm) and stable parameters (COV < 20%), and were therefore selected for ULM reconstruction. All ULM parameters showed high inter-operator agreement (ICC 0.876-0.988). Based on the established ULM framework, higher intratumoral mean curvature was independently associated with poor differentiation [area under the curve (AUC) 0.91], and higher peritumoral mean curvature independently predicted MVI status (AUC 0.78).
The reproducible ULM framework enables stable, high-resolution microvascular imaging of HCC. ULM-derived parameters hold potential as novel biomarkers for predicting differentiation grade and MVI status of HCC.
Participants were prospectively enrolled and underwent contrast-enhanced ultrasound and ULM imaging. We first compared frame selection strategies by calculating coefficients of variation (COVs) of ULM parameters derived from motion-variance curve (MVC) alone versus MVC combined with time-intensity curve (TIC). ULM resolution and parameter stability were then assessed across low (0-399), medium (400-699), and high (700-1000) frame counts. Regions of interest (ROIs) were manually drawn on grayscale ultrasound and mapped to ULM images. Inter-operator agreement was evaluated using intraclass correlation coefficients (ICCs). Participants were grouped by pathological differentiation and MVI status. The predictive performance of ULM parameters was assessed with multivariable logistic regression.
Sixty-one HCC participants were enrolled (11 poorly differentiated, 50 well differentiated; 30 MVI-positive, 31 MVI-negative). The MVC+TIC strategy yielded significantly lower COVs, indicating higher repeatability of ULM parameters. Medium frame counts (400-699) provided high resolution (minimum vessel diameter 91.3 ± 22.7 μm) and stable parameters (COV < 20%), and were therefore selected for ULM reconstruction. All ULM parameters showed high inter-operator agreement (ICC 0.876-0.988). Based on the established ULM framework, higher intratumoral mean curvature was independently associated with poor differentiation [area under the curve (AUC) 0.91], and higher peritumoral mean curvature independently predicted MVI status (AUC 0.78).
The reproducible ULM framework enables stable, high-resolution microvascular imaging of HCC. ULM-derived parameters hold potential as novel biomarkers for predicting differentiation grade and MVI status of HCC.