An accurate, straightforward computer vision algorithm for optimal tumor-feeding visualization in cone-beam computed tomography hepatic arteriography: A preliminary study
Bibliographic record
Abstract
AIM: Although standardized 3D volume rendering techniques (VRT) and embolization guidance visualize and identify tumor-feeding arteries, current vessel tracking software lacks automatic angle recommendations. This forces an operator, e.g. an interventional radiologist, to leave an ongoing procedure to manually manipulate the system and find the best angle for each feeding vessel-requiring time-consuming re-scrubbing. We propose a computer vision algorithm that suggests a rotation/angle in the VRT where a tumor-feeding artery's view is maximized. We focus on hepatocellular carcinoma. METHODS: Our algorithm accepts a series of post-embolization guidance frames extracted from the 3D VRT; the VRT is rotated in 5° intervals from, e.g., ±15°, fixing one axis (e.g. CRAN/CAUD) and rotating the other (e.g. LAO/RAO). Our algorithm segments the embolization guidance line and recommends 4 views/angles by maximizing the features of line length (contour area) and convex hull area. We developed/iterated our algorithm using 19 patient cases and feedback from various experts. RESULTS: Over a 50-patient internal validation set, according to an interventional radiologist with 33 years of experience, a view/angle sufficient for the embolization task was always present among the top-4 views/angles suggested by our algorithm (100% retrieval relevance). CONCLUSION: Sufficient view/angle selection for hepatic artery embolization can be achieved using traditional computer vision. Our technique is much faster and more explainable than deep learning approaches, and could greatly improve radiologists' procedural efficiency. We recommend conducting a larger study with more patients and further technical iteration.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".