Scaling Tumor Segmentation: Best Lessons from Real and Synthetic Data

2025-10-16

AI Summary

Voxel-level annotation scarcity and heavy reliance on expert radiologists hinder abdominal multi-organ tumor segmentation. Method: We propose a real-synthetic data co-training paradigm and introduce AbdomenAtlas 2.0—the largest fine-grained annotated abdominal CT dataset to date (10,135 cases, 15,130 tumor instances), fully voxel-labeled by 23 radiologists. We rigorously validate the strong generalization gains of synthetic data in few-shot settings and establish a data-efficient scaling pathway. Results: Our approach achieves +7% Dice improvement on in-distribution test sets and +16% on out-of-distribution sets, substantially outperforming prior public benchmarks. AbdomenAtlas 2.0 establishes a high-quality, reproducible benchmark for cross-organ tumor segmentation and pioneers a novel data-algorithm co-development paradigm grounded in scalable, expert-validated annotation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-instance/Multi-view Learning
• Data Mining & Knowledge Management: Mining of Visual, Multimedia & Multimodal Data

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

AI for tumor segmentation is limited by the lack of large, voxel-wise annotated datasets, which are hard to create and require medical experts. In our proprietary JHH dataset of 3,000 annotated pancreatic tumor scans, we found that AI performance stopped improving after 1,500 scans. With synthetic data, we reached the same performance using only 500 real scans. This finding suggests that synthetic data can steepen data scaling laws, enabling more efficient model training than real data alone. Motivated by these lessons, we created AbdomenAtlas 2.0--a dataset of 10,135 CT scans with a total of 15,130 tumor instances per-voxel manually annotated in six organs (pancreas, liver, kidney, colon, esophagus, and uterus) and 5,893 control scans. Annotated by 23 expert radiologists, it is several orders of magnitude larger than existing public tumor datasets. While we continue expanding the dataset, the current version of AbdomenAtlas 2.0 already provides a strong foundation--based on lessons from the JHH dataset--for training AI to segment tumors in six organs. It achieves notable improvements over public datasets, with a +7% DSC gain on in-distribution tests and +16% on out-of-distribution tests.

Problem

• AI tumor segmentation limited by scarce annotated medical datasets
• Synthetic data enhances training efficiency compared to real data
• AbdomenAtlas 2.0 provides large-scale annotated dataset for multi-organ tumor segmentation

Innovation

• Combining real and synthetic data for tumor segmentation
• Creating AbdomenAtlas 2.0 with 10,135 annotated CT scans
• Achieving significant performance gains in distribution tests