SMC-LUD: Liver Ultrasound Images for HCC vs Hemangioma Classification

5,466 anonymized 2D B-mode ultrasound images categorized into hepatocellular carcinoma (HCC) and hemangioma classes. The dataset includes a metadata CSV with patient_id, image_filename, and class_label, and is organized by patient subdirectories. It was released by Jisoo Tak from Samsung Medical Center in April 2026 under a CC-BY license.

Use Cases

• Train binary classification models for HCC vs hemangioma based on 2D ultrasound images.
• Benchmark model performance using the provided training, validation, and test splits.
• Analyze patient-level tumor characteristics based on the demographic and tumor metadata.
• Develop models that leverage patient-specific groupings of multiple ultrasound images.

Strengths

• Contains 5,466 ultrasound images, providing a substantial volume for training.
• Includes detailed patient-level metadata such as age, sex, and tumor stage for 1,021 patients.
• Provides a structured split of 1,858 training, 530 validation, and 306 test images for the 'Clean' subgroup.

Limitations

• Row count for the metadata CSV is unknown, which may limit suitability assessment.
• Column-level documentation for the metadata is absent; field semantics must be inferred after download.
• Data is restricted to non-commercial research use only, limiting commercial applications.

Provenance

Source

Samsung Medical Center (SMC)

Collection Method

Likely collected from clinical ultrasound examinations.

Freshness

Last updated 2026-04-27 07:01:17; freshness should be verified.