Hybrid Context Forests and Bilateral Viterbi Matching for 3D Rib Centerline Tracing in CT

Quan Wang

eess.IVOct 12, 2026

Hybrid Context Forests and Bilateral Viterbi Matching for 3D Rib Centerline Tracing in CT

Authors: Quan Wang

Abstract

Automatic extraction and anatomical numbering (R01-R12, L01-L12) of all 24 rib centerlines in thoracic CT is essential for unfolded rib visualization, fracture detection, and surgical navigation, but remains challenging due to adjacent-bone confluence, calcified costal cartilage bridges, kyphotic pose variation, and partial visibility of terminal ribs. We present ribline, an open-source three-stage framework combining: (1) an automatically detected lung-apex and vertebral-midline coordinate frame with a kyphosis-aware azimuthal angle for point ordering along each rib; (2) a hybrid voxel detector fusing an anatomically anchored Semantic Context Forest (SCF) with a lightweight 3D U-Net; and (3) a global stage that decomposes the probability skeleton into candidate curves, assigns rib indices via a bilateral Viterbi dynamic program coupled across both hemithoraces, and refines curves via template recovery and normal-plane centering (RibSnake). On the Siemens Corporate Research 2 mm thoracic CT benchmark (47 training and 12 evaluation scans), our pipeline achieves a labeled Chamfer distance of 2.05 mm, labeled clDice of 0.886 (3 mm) and 0.945 (4 mm), 97.2% labeling accuracy (278/286 ribs), and zero false-positive curves on the evaluation split, and 3.64 mm (median 2.39 mm) with 91.8% labeling accuracy across all 59 annotated scans (1,402 ribs) in 5-fold cross-validation. Code: https://github.com/wq2012/ribline Pretrained Models: https://huggingface.co/collections/wq2012/ribline

Figures & tables

Figure 1: Architecture of the proposed three-stage ribline framework: Stage 1 standardizes Hounsfield units and establishes the lung-apex / vertebral-midline coordinate frame; Stage 2 fuses an anatomically anchored Semantic Context Forest (SCF) with a lightweight 3D U-Net; Stage 3 extracts the skeleton graph, solves contralateral-coupled bilateral Viterbi labeling, and refines centerlines via template recovery and RibSnake normal-plane centering.

Figure 2: Canonical thoracic reference frame and statistical shape prior. (a) Coronal CT MIP with automatically detected lung apex $Z_{\mathrm{apex}}$ , vertebral midline trajectory $X_{\mathrm{spine}}(Z)$ , and ground-truth rib centerlines. (b) The 24-rib canonical statistical atlas $\bar{\mathcal{C}}_{s,k}$ fitted on the 47 training scans.

Figure 3: Coronal MIPs of centerline probability maps on evaluation scan pet_ct_bc__eh_Anonymous_0004 : (a) CT, (b) Pass-1 forest, (c) anatomically anchored SCF, (d) 3D U-Net, (e) hybrid max-fusion, (f) dilated ground truth.

Method	LCD (mm) $\downarrow$	95 % CI	UCD $\downarrow$	lab. clDice $\uparrow$		Labeling accuracy (%) $\uparrow$				Recall $\uparrow$	Prec. (FP) $\uparrow$	NLD
Method	mean $\pm$ SD		(mm)	3 mm	4 mm	$\mathrm{Acc}_{A}$	$\mathrm{Acc}_{F}$	$\mathrm{Acc}_{I}$	$\mathrm{Acc}_{T}$	(%)	(%)	$\downarrow$
RF Pass-1 + Z-sort	$14.10\pm 11.10$	$[7.96,20.46]$	$3.31$	$0.548$	$0.606$	$64.3$	$79.2$	$62.9$	$63.6$	$91.3$	$95.5$ (13)	$0.264$
SCF Pass-2 + Z-sort	$21.30\pm 13.33$	$[13.79,28.67]$	$4.16$	$0.416$	$0.452$	$43.0$	$45.8$	$44.2$	$27.3$	$76.2$	$100.0$ (0)	$0.254$
SCF Pass-2 + Stage 3	$9.36\pm 10.57$	$[4.24,16.19]$	$3.73$	$0.683$	$0.720$	$71.0$	$54.2$	$73.8$	$59.1$	$78.0$	$99.6$ (1)	$0.226$
Anatomical SCF + Stage 3	$2.88\pm 1.32$	$[2.23,3.73]$	$2.41$	$0.845$	$0.915$	$95.1$	$91.7$	$95.0$	$100.0$	$95.1$	$99.7$ (1)	$0.088$
3D U-Net + Z-sort	$7.57\pm 7.43$	$[3.97,12.17]$	$2.75$	$0.710$	$0.813$	$83.6$	$70.8$	$86.7$	$63.6$	$92.0$	$100.0$ (0)	$0.147$

Table 1: Ablation on the 12 evaluation scans (

286

GT ribs, RIB_007 annotation fixed). LCD: labeled symmetric Chamfer distance with 50 mm penalty for missed labels (subject mean

\pm

SD and 95 % bootstrap CI); clDice: labeled, 3 mm / 4 mm; accuracies, recall and precision pooled over ribs (symmetric 10 mm criterion); FP: false-positive curves out of all predicted curves. Note the protocol caveat: Stage-3 constants were developed on these scans.

Figure 4: Three evaluation scans (dashed gray: ground truth; colored: predicted labels 1–12). Top: SCF Pass-2 + Z-sort baseline. Bottom: hybrid + full Stage 3; titles give the number of correctly labeled ribs under the symmetric 10 mm criterion and the LCD.

Figure 5: Per-rib analysis on the 12 evaluation scans: (a) LCD per rib pair (log scale, missing labels

=50

mm); (b) labeling accuracy per rib pair; (c) pooled

\mathrm{Acc}_{A}

\mathrm{Acc}_{F}

and labeled clDice@4 mm with 95 % bootstrap CIs.

Method	LCD (mm) $\downarrow$	95 % CI	UCD (mm) $\downarrow$	lab. clDice $\uparrow$		$\mathrm{Acc}_{A}\uparrow$	$\mathrm{Acc}_{F}\uparrow$	$\mathrm{Acc}_{I}\uparrow$	$\mathrm{Acc}_{T}\uparrow$	Recall $\uparrow$	Prec. (FP) $\uparrow$	NLD $\downarrow$
Anatomical SCF + Stage 3	$3.62\pm 3.97$	$[2.67,4.91]$	$2.97$	$0.830$	$0.895$	$92.4$	$89.2$	$92.4$	$95.7$	$94.8$	$97.4$ (29)	$0.094$
3D U-Net + Stage 3	$5.84\pm 7.20$	$[3.95,8.03]$	$2.98$	$0.785$	$0.826$	$83.9$	$69.9$	$86.0$	$76.3$	$91.6$	$97.8$ (25)	$0.142$
Hybrid + Viterbi (w/o RibSnake)	$2.98\pm 2.11$	$[2.42,3.63]$	$2.77$	$0.875$	$0.930$	$\mathbf{95.6}$	$\mathbf{88.2}$	$\mathbf{96.1}$	$\mathbf{97.8}$	$\mathbf{96.1}$	$\mathbf{98.5}$ (17)	$0.155$
Hybrid + full Stage 3 (ours)	$\mathbf{2.63\pm 1.91}$	$[2.14,3.23]$	$\mathbf{2.62}$	$\mathbf{0.885}$	$\mathbf{0.932}$	$95.4$	$86.0$	$\mathbf{96.1}$	$\mathbf{97.8}$	$95.6$	$98.2$ (20)	$\mathbf{0.083}$

Table 2: Frozen pipeline (released detectors, atlas, and Stage-3 constants) on the 47 annotated training scans (

1{,}116

GT ribs; detectors in-sample, Stage-3 constants never tuned on these scans). Same metric definitions as Table 1 .

Method	LCD (mm) $\downarrow$	95 % CI	UCD (mm) $\downarrow$	lab. clDice $\uparrow$		$\mathrm{Acc}_{A}\uparrow$	$\mathrm{Acc}_{F}\uparrow$	$\mathrm{Acc}_{I}\uparrow$	$\mathrm{Acc}_{T}\uparrow$	Recall $\uparrow$	Prec. (FP) $\uparrow$	NLD $\downarrow$
Anatomical SCF + Stage 3	$5.28\pm 4.62$	$[4.16,6.56]$	$3.44$	$0.769$	$0.840$	$86.9$	$72.6$	$87.7$	$93.0$	$90.8$	$95.3$ (67)	$0.126$
3D U-Net + Stage 3	$7.44\pm 8.93$	$[5.31,9.85]$	$3.30$	$0.745$	$0.785$	$78.2$	$61.5$	$80.9$	$67.8$	$87.0$	$\mathbf{98.1}$ (26)	$0.170$
Hybrid + Viterbi (w/o RibSnake)	$4.17\pm 4.01$	$[3.29,5.28]$	$3.07$	$0.807$	$0.879$	$91.4$	$81.2$	$\mathbf{92.2}$	$93.9$	$94.3$	$97.9$ (29)	$0.195$
Hybrid + full Stage 3 (ours)	$\mathbf{3.64\pm 3.90}$	$[2.79,4.74]$	$\mathbf{2.85}$	$\mathbf{0.830}$	$\mathbf{0.889}$	$\mathbf{91.8}$	$\mathbf{82.9}$	$\mathbf{92.2}$	$\mathbf{96.5}$	$\mathbf{94.6}$	$97.7$ (33)	$\mathbf{0.097}$

Table 3: 5-fold cross-validation across all 59 annotated SCR scans (

1{,}402

GT ribs; per-fold re-fitted atlas, Anatomical SCF, and 3D U-Net; frozen Stage-3 constants; benchmark_cv5.json ).

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

Module / Script	Section	Implemented Functionality
ribline.data.image_reader	Sec. 2	Analyze 7.5 ( .hdr/.img ), NIfTI ( .nii/.nii.gz ), and MetaImage ( .mha/.mhd ) 3D I/O with spacing and origin preservation.
ribline.data.centerline_parser	Sec. 5	Parsing of ground-truth ( _interp.xls ) and predicted 3D centerline coordinate files.
ribline.model.anatomy	Sec. 2	Hounsfield standardization ( standardize_hu ), lung apex and spine midline detection, kyphosis-aware azimuthal angle, and 24-rib statistical atlas ( RibCageStatisticalAtlas ).
ribline.model.features	Sec. 3	Multi-scale 3D Gaussian derivatives, diagonal-curvature tubularity proxy, canonical coordinates, and Random Shift Intensity/Probability Differences (RSID).
ribline.model.forest	Sec. 3	Balanced random forest wrappers ( RibForest ) for Pass-1, Pass-2 auto-context, and anatomically anchored SCF detectors.
ribline.model.unet.model	Sec. 3	Plain 3-level 3D U-Net ( RibUNet3D , $339{,}889$ parameters) with GroupNorm and sliding-window inference.

Appendix

Table 4: Mapping between modules in the open-source ribline library ( https://github.com/wq2012/ribline ) and the sections of this paper.

Hugging Face Repository ID	Model Variant & Included Checkpoints	Size	12-Scan LCD	12-Scan Acc	5-Fold CV LCD / Acc
wq2012/ribline-3d-ct-centerline	Full Hybrid Suite (SCF + 3D U-Net + Pass 1/2 + Atlas)	$529$ MB	$\mathbf{2.05\pm 0.48}$ mm	$\mathbf{97.2\,\%}$	$\mathbf{3.64\pm 3.90}$ mm / $\mathbf{91.8\,\%}$
wq2012/ribline-anatomical-scf	Standalone Anatomically Anchored SCF + Atlas (CPU)	$152$ MB	$2.88\pm 1.32$ mm	$95.1\,\%$	$5.28\pm 4.62$ mm / $86.9\,\%$
wq2012/ribline-3d-unet	Standalone Lightweight 3D U-Net ( $340$ k params) + Atlas	$1.7$ MB	$3.40\pm 1.99$ mm	$92.0\,\%$	$7.44\pm 8.93$ mm / $78.2\,\%$
wq2012/ribline-scf-pass2-autocontext	Two-Pass Auto-Context SCF Baseline + Atlas	$377$ MB	$9.36\pm 10.57$ mm	$71.0\,\%$	—

Appendix

Table 5: Pretrained model repositories published on Hugging Face ( https://huggingface.co/collections/wq2012/ribline ). All checkpoints are trained on the 47 SCR training scans; 12-scan and 59-scan 5-fold CV metrics use the symmetric 10 mm criterion and 50 mm missed-rib LCD penalty.

3D Medical Image Segmentation

3D Medical Image Segmentation identifies organs, lesions, and fine structures in volumetric scans for measurement and clinical planning.

See all 212 papers

RibAssist 3D: Biplanar Rib-Fracture Detection, Addressing, and Selective 3D Localization from CT-Derived Projections
Aug 7, 2026Kabila Haile SobokaSelective Prediction Medical Image Analysis
The University of Texas at Austin Austin, Texas, USA
VesselFusion: Diffusion Models for Vessel Centerline Extraction from 3D CT Images
Mar 9, 2026Soichi Mita, Shumpei Takezaki, Ryoma Bise3D Medical Imaging 3D Diffusion Models
Kyushu University, Fukuoka, Japan · German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany
ARC-CT: Anatomy-Routed Contrastive Vision-Language Learning for 3D Chest CT
Aug 28, 2026Huseyin Umut Isik, Mehmet Alp Ozaydin, Sila Kurugol +1Radiology VLMs Medical Image Classification
Department of Computer Engineering, METU, Ankara, Turkey · Quantitative Intelligent Imaging Lab, Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA · METU-DTX Digital Transformation and Innovation Center, Ankara, Turkey

Aug 7, 2026J/K move · Enter open · S save

RibAssist 3D: Biplanar Rib-Fracture Detection, Addressing, and Selective 3D Localization from CT-Derived Projections

Kabila Haile Soboka

The University of Texas at Austin Austin, Texas, USA

Selective Prediction Medical Image Analysis Cross-View Image Matching Computed Tomography

Open paper AskPDF arXiv

Rib fractures are common and time-consuming to localize on computed tomography (CT). We ask whether fractures detected independently in two orthogonal CT-derived projections (anteroposterior and lateral) can be paired across views and triangulated into reliable 3D points at a controlled rate of false outputs, and we answer it with a staged diagnostic study. The projection geometry is exact, and given correct correspondence, localization is accurate (median 4.0 mm, 88% within 10 mm, 93.6% rib-exact). On a sealed 55-case cohort, a large share of fractures is in principle recoverable (61.1% dual-view availability, and a correct pair present in the candidate graph for 58.4% of fractures), yet the binding limitation is neither geometry nor localization but confidence-limited cross-view correspondence. A controlled detector-by-correspondence factorial attributes the operational gain to lateral-detector quality rather than the tested matching methods; retraining the lateral detector produces the first nonzero controlled-budget reconstructions. Under a deliberately conservative commitment policy, a pre-specified sealed pass promotes 15 of 601 fractures to correct 3D localizations at 0.436 false points per case (2.50% end-to-end commitment yield), and committed points are accurate (median 1.49 mm, 93% rib-exact). The low yield is a consequence of confidence-gated abstention, not of geometry or detection: the study establishes a reproducible framework for selective 3D localization and identifies cross-view correspondence as the dominant operational bottleneck.

Abstract

3D Medical Image Segmentation

Explore similar work

RibAssist 3D: Biplanar Rib-Fracture Detection, Addressing, and Selective 3D Localization from CT-Derived Projections

VesselFusion: Diffusion Models for Vessel Centerline Extraction from 3D CT Images

ARC-CT: Anatomy-Routed Contrastive Vision-Language Learning for 3D Chest CT