AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

AbdomenAtlas-8K: Annotating 8,000 CT Volumes for Multi-Organ Segmentation in Three Weeks

Chongyu Qu, Tiezheng Zhang, Hualin Qiao, Jie Liu, Yucheng Tang, Alan Yuille, Zongwei Zhou

NeurIPS 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Interactive and referring segmentation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train segmentation model on AbdomenAtlas-8K instead of JHH private data (SwinUNETR)SwinUNETR · JHH testmDSC (%): 84.8 -> 86.5Improved
train on AbdomenAtlas-8K instead of JHH (UNETR)UNETR · JHH testmDSC (%): 78.6 -> 86.6Improved
train on AbdomenAtlas-8K instead of JHH (U-Net)U-Net · JHH testmDSC (%): 84.8 -> 87.3Improved
train on AbdomenAtlas-8K instead of JHH (SegResNet)SegResNet · JHH testmDSC (%): 87.6 -> 87.0No clear effect
train on AbdomenAtlas-8K instead of JHH (SwinUNETR)SwinUNETR · JHH testmNSD (%): 66.5 -> 60.8Hurt
train on AbdomenAtlas-8K instead of JHH (UNETR)UNETR · JHH testmNSD (%): 53.9 -> 59.5Improved
train on AbdomenAtlas-8K instead of JHH (U-Net)U-Net · JHH testmNSD (%): 64.2 -> 61.0Hurt
train on AbdomenAtlas-8K instead of JHH (SegResNet)SegResNet · JHH testmNSD (%): 64.9 -> 60.4Hurt
fine-tune with annotator-revised labels over active learning rounds (aorta) (ablation)Active learning pipeline · JHH 1000 CTsDSC (%): 72.3 -> 83.7Improved
fine-tune with annotator-revised labels over active learning rounds (postcava) (ablation)Active learning pipeline · JHH 1000 CTsDSC (%): 76.1 -> 78.6Improved
prioritize top 5% of CTs by attention map for human revision instead of annotating every organ manuallyActive learning annotation · 8,000 CT volumesannotation time: 30.8 years -> three weeksImproved
manual per-CT annotation to attention-guided revisionActive learning annotation · per CT volumeannotation speedup: 60×8/15=32Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy