ReXGroundingCT

ReXGroundingCT is a large-scale 3D chest CT dataset linking free-text radiology findings to pixel-level segmentations in volumetric imaging. It comprises 3,142 non-contrast chest CT scans with 8,028 annotated findings (16,301 entities) from the CT-RATE dataset. ReXGroundingCT enables sentence-level grounding for both focal and non-focal lung and pleural abnormalities across 14 categories. On ReXrank, we are hosting ReXGroundingCT's testset, which contains 100 CT scans with exhaustive radiologist annotations for all visible findings.

---

Global Dice: Average Dice per finding per case

Global HIT Rate: Proportion of findings that have Dice >= 0.1

Instance Precision: TP / (TP + FP), where True Positives are instances that have a Dice >= 0.2

Instance Recall: TP / (TP + FN)

---

Model	Global Dice	Global HIT Rate	Instance Precision	Instance Recall	Instance F1
DAGG 2/23/2026 Submission 2 Submission 1 Submission 2	0.253	0.517	0.211	0.299	0.247
VoxTell German Cancer Research Center (DKFZ) Heidelberg University 3/24/2026	0.285	0.615	0.185	0.294	0.227
SAT-FT Shanghai Jiao Tong University	0.205	0.473	0.062	0.383	0.107
BiomedParseV2 Microsoft	0.025	0.066	0.006	0.071	0.012
SegVol Shanghai Jiao Tong University Beijing Academy of Artificial Intelligence	0.000	0.000	0.000	0.000	0.000
SAT Shanghai Jiao Tong University	0.000	0.000	0.000	0.000	0.000

---

Category	n	Dice	Hit Rate

Legend: Category codes starting with 1 are typically non-focal lung/airway/pleural abnormalities; codes starting with 2 are typically focal lung/airway/pleural opacities.