Abdominal CT Scans with 3D Lesion Annotations

Over 6,000 contrast-enhanced abdominal CT scans curated from the public DeepLesion dataset, focusing on studies with visible lesions or anomalies. The dataset was created by author 'mehiii' for the 3DLAND project, a benchmark for organ-aware 3D lesion segmentation. It was last updated on the platform in February 2026.

Use Cases

• Train 3D convolutional neural networks for pixel-wise lesion segmentation using the provided 3D scan volumes and segmentation masks.
• Develop organ-aware anomaly localization models by leveraging the structured, organ-specific annotations within the 3D volumes.
• Benchmark the performance of segmentation algorithms on a large-scale collection of abdominal CT scans containing annotated lesions.

Strengths

• Contains over 6,000 contrast-enhanced abdominal CT scans, providing a substantial volume of data.
• Scans are specifically curated to include visible lesions or anomalies, ensuring relevance for pathology-focused tasks.

Limitations

• Dataset size, specific row/column counts, and file formats are not detailed in the provided description.
• The geographic and temporal coverage of the source DeepLesion scans is unspecified, which may introduce demographic or technical bias.

Provenance

Source

Curated from the publicly available DeepLesion dataset.

Collection Method

Scans were selected from DeepLesion based on the presence of visible abdominal lesions, then transformed into a structured, organ-aware 3D segmentation benchmark.

Freshness

Last updated on the platform in February 2026.