Dual-Branch Vascular Segmentation Model Based on Anisotropic Attention
Bibliographic record
Abstract
Vascular segmentation is significant for diagnosing and treating vascular diseases. However, because of the fuzzy boundary of vessels, the variable shape of diseased vessels, and the significant differences between different samples, the segmentation model should accurately determine the differences between vessels and background classes and analyze the connectivity within vessels. This study proposes a novel three-dimensional vascular segmentation network, CAU-Net, based on centerline constraints and anisotropic attention. In response to the difficulties in vascular segmentation, the basic network structure, ResU-Net, is improved to construct an anisotropic attention module. This module extracts vascular spatial anisotropic features from three directions based on the unique spatial anisotropy of the vascular structure and models the correlation between feature channels to learn the three-dimensional spatial information of the vessels. By using the main auxiliary dual-branch model, b-Net performs semantic segmentation on vessels, whereas a-Net learns the continuity features of vessel centerlines, constrains the vascular segmentation results of b-Net, and ensures the integrity of the vascular segmentation results. The experimental results on the publicly available dataset 3D-IRCADb-01 shows that for the segmentation of portal and hepatic veins, CAU-Net achieves Dice coefficients of (74.80±8.05)% and (76.14±6.89)%, NSD coefficients of(54.80±8.09)% and(50.40±5.22)%, clDice coefficients of (72.43±8.26)% and(70.84±6.05)%, Branch Detection(BD) rates of (46.47±12.89)% and(39.19±7.97)%, and Tree length Detection(TD) rates of(67.08±15.59)% and(61.47±9.32)%, respectively. Component ablation experiments are conducted on the publicly available cerebrovascular dataset IXI, and the average Dice, NSD, clDice, BD, and TD values of the model on the validation set are(94.11±0.39)%, (96.53±0.37)%, (95.83±0.59)%, (98.64±1.63)%, and(95.44±1.22)%, respectively. Compared to the Baseline, the average Dice, NSD, clDice, BD, and TD values of the proposed model increased by 0.92%, 0.82%, 0.92%, 1.11%, and 1.60%, respectively. The CAU-Net vascular segmentation model can significantly improve the accuracy and completeness of vascular segmentation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.009 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".