3DLAND

The First Large-Scale Organ-Aware 3D Lesion Benchmark

Empowering Medical AI with over 20,000+ high-fidelity 3D annotations across 7 abdominal organs, fully validated by expert radiologists.

"
Under Review at ICLR 2026
"
Qualitative Segmentation Results and Prompt Comparison

6,000+

CT Volumes

20,000+

3D Lesion Masks

7 Organs

Multi-Organ Coverage

1295

Unique Patient


Comparison with State-of-the-Art Benchmarks

3DLAND establishes a new standard for abdominal anomaly localization. The charts below illustrate the order-of-magnitude increase in the number of annotated 3D lesions compared to the previous largest public datasets (e.g., LiTS, KiTS, MSD). While prior works often focused on single organs, 3DLAND provides comprehensive multi-organ coverage, including the first-ever public benchmark for Gallbladder lesions [cite: 145] and significantly expanding data availability for the Stomach and Spleen.

Abstract

Existing medical imaging datasets for abdominal CT often lack three-dimensional annotations, multi-organ coverage, or precise lesion-to-organ associations. To address this gap, we introduce 3DLAND, a large-scale benchmark dataset comprising over 6,000 contrast-enhanced CT volumes with over 20,000 high-fidelity 3D lesion annotations linked to seven abdominal organs: liver, kidneys, pancreas, spleen, stomach, and gallbladder.

Our streamlined three-phase pipeline integrates automated spatial reasoning, prompt-optimized 2D segmentation, and memory-guided 3D propagation, validated by expert radiologists with surface dice scores exceeding 0.75. By providing diverse lesion types and patient demographics, 3DLAND enables scalable evaluation of anomaly detection, localization, and cross-organ transfer learning.

The 3DLAND Pipeline

We introduce a streamlined Three-Phase Pipeline to transform 2D bounding boxes into high-fidelity 3D masks.

3DLAND Three-Phase Pipeline Architecture

Phase I

Lesion-to-Organ Assignment via Spatial Reasoning (MONAI).

Phase II

Prompt-Optimized 2D Segmentation (MedSAM1 + Center Point).

Phase III

Memory-Guided 3D Propagation (MedSAM2).

Qualitative Results

Comparison of segmentation performance using different prompting strategies on real-world clinical cases. Our method (Blue: BBox + Center Point) achieves the most precise boundaries compared to other strategies, reducing over-segmentation and improving localization accuracy.

Qualitative Segmentation Results and Prompt Comparison

BibTeX

@inproceedings{3dland2026,
  title={3DLAND: 3D Lesion Abdominal Anomaly Localization Dataset},
  author={Anonymous},
  booktitle={Under Review at ICLR 2026},
  year={2026}
}