A public-data reproducibility benchmark of a 2.5-D U-Net with a slice-level topology term for type-B aortic dissection segmentation
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Type-B aortic dissection requires separation of the true and false lumina in heterogeneous, anisotropic CTA. A topology term may reduce fragmentation, but its reproducibility must be tested under a fixed case split and independent native-grid evaluation.
Methods
We used the public Graz type-B aortic dissection CTA collection (40 cases; CC BY 4.0) and froze a 28/6/6 case split for training, development, and held-out testing. A compact 2.5-D U-Net trained with cross-entropy plus multiclass Dice was compared with the identical model augmented by a fixed-weight (0.5) differentiable slice-level soft clDice term. Two training seeds were run for each arm. All 40 cases passed the NRRD audit; predictions for six held-out cases were restored to their native grids and scored with case-level 3-D voxel-grid Dice, voxel-grid morphological clDice, 6-connected component counts, physical-volume error, and 2,000-resample case-level paired bootstrap intervals.
Results
With seed 20261007, topology-minus-baseline differences were −0.028 (95% CI −0.053 to −0.004) for true-lumen Dice, +0.032 (−0.006 to +0.096) for true-lumen voxel-grid clDice, +0.071 (+0.037 to +0.103) for false-lumen Dice, and +0.073 (+0.048 to +0.103) for false-lumen voxel-grid clDice. With seed 20261008, the corresponding differences were −0.040 (−0.065 to −0.018), +0.013 (−0.034 to +0.071), −0.099 (−0.187 to −0.004), and −0.058 (−0.124 to −0.001). Descriptive pooled means over 12 case rows per arm were true-lumen Dice 0.577 for topology versus 0.611 for baseline and false-lumen Dice 0.409 versus 0.422; these pooled rows are not independent cases.
Conclusions
Under this fixed six-case test split, the corrected slice-level topology term did not produce a consistent benefit. True-lumen Dice was lower in both seeds, while false-lumen effects reversed direction between seeds; topology-clDice changes were small and/or interval-crossing. The benchmark supports an auditable failure-analysis workflow but does not support a general topology benefit, clinical use, external generalization, or a panvascular decision system. The 3-D clDice quantity is an evaluation descriptor on the voxel grid; the training term is slice-level and does not enforce physical 3-D connectivity.
Current executed scope and claim boundary
This executed benchmark is restricted to the 40-case Graz type-B aortic dissection CTA collection with expert true- and false-lumen annotations. All 40 cases passed the audit; the frozen case split contains 28 training, 6 development, and 6 held-out test cases. Four runs (baseline/topology arms with seeds 20261007 and 20261008) produced native-grid predictions for all six held-out cases. The current estimand is descriptive case-level segmentation performance within this single public source. AVT, AortaSeg-60, ImageTBAD, and the ancillary multi-structure CCTA collection are documented as possible future external resources, but they are not part of the current endpoint, training set, or performance claim. This is a proof-of-concept segmentation benchmark, not a panvascular clinical system. The present study does not estimate multicentre generalisation, perform a TEVAR landing-zone assessment, evaluate device fit, or use follow-up outcomes. Any virtual-TEVAR geometry extension remains separately gated future work.
