DSAD: Laparoscopic Surgical Anatomy Images with Multi-Annotator Masks

13,195 laparoscopic images from 32 robot-assisted rectal resection surgeries, each annotated for semantic segmentation of eleven abdominal anatomical structures. The dataset was created by nct-tso and last updated on March 30, 2026. Each image's annotation was performed by three independent annotators and subsequently reviewed.

Use Cases

• Training semantic segmentation models for abdominal anatomy based on the provided masks for 11 structures.
• Developing surgical navigation or augmented reality systems based on real-time anatomical identification in laparoscopic video.
• Benchmarking model performance for multi-class segmentation in complex, real-world surgical environments.
• Studying inter-annotator agreement and annotation quality for medical imaging tasks.

Strengths

• 13,195 images provide a substantial base for training computer vision models.
• Annotations for eleven distinct anatomical structures enable detailed surgical scene analysis.
• Each image was annotated by three independent annotators, which likely improves label reliability.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific model training needs.

Provenance

Source

huggingface, author nct-tso

Collection Method

Images collected from 32 robot-assisted rectal resection surgeries.

Time Range

null

Freshness

Last updated 2026-03-30 13:55:03; freshness should be verified.

Geography

null