Contrast-enhanced ultrasound (CEUS) substantially improves the potential of ultrasound (US) for the identification and characterization of focal liver lesions (FLLs). Compared to contrasted-enhanced MRI and CT, it has some unique advantages, such as the absence of ionizing radiation, and easy operability and repeatability. However, the efficacy of CEUS in diagnosing liver lesions is challenged by several factors including being highly dependent on doctor's experience, low signal-to-noise ratio, and low interobserver agreement. Therefore, it is a beneficial attempt to construct an intelligent CEUS diagnosis system using digital information technology. This study aims to collect standard data of CEUS cines recordings and develop deep learning model for accurate segmentation, detection and classification of liver lesions.

Inclusion Criteria:

1. patients with a solid liver tumor visible during routine ultrasound and received CEUS.
2. disease history and standard of reference of the lesions can be acquired

Exclusion Criteria:

1. hypersensitivity for ultrasound contrast media
2. pregnant or lactating patients
3. previously treated lesions or local relapse from previously treated lesions
4. diffuse tumors

patients with focal liver lesions