Data were generated as part of two ongoing clinical trials investigating the use of contrast-enhanced ultrasound to a) characterize indeterminate liver lesions and b) monitor treatment response to loco regional therapy. Ultrasound data was obtained on a variety of state of the art ultrasound scanners with curvilinear probes. Gain, dynamic range, focus position and depth were optimized for image quality by the performing sonographer. Images of the mass in both sagittal and transverse planes were obtained and saved in DICOM format. Full cine-loops of the contrast enhanced ultrasound are also saved in DICOM format. The reference standard used for lesion characterization included tissue pathology and contrast-enhanced cross-sectional imaging within 1 month of the ultrasound exam. The initial treatment response to transarterial chemoembolization is also available for many hepatocellular carcinoma (HCC) cases and uses pathology, retreatment angiography, or longer-term tumor response on cross-sectional imaging as a reference standard.

We expect these images can be used in a wide variety of image processing application. We are currently exploring a variety of automated intelligence algorithms for AI-based lesion characterization. Algorithms for object detection and segmentation may also be of interest. As the studies that have generated this data are also ongoing, it is expected that we can add volumetric data, contrast-enhanced ultrasound cine loops, and longer-term treatment response data to this data set in the future.