Using uncertainty quantification to improve reliability of video-based skill assessment metrics in central venous catheterization
Bibliographic record
Abstract
Computed-based skill assessment relies on accurate metrics to provide comprehensive feedback to trainees. Improving the accuracy of video-based metrics computed using object detection is generally done by improving the performance of the object detection network, however increasing its performance requires resources that cannot always be obtained. This study aims to improve the accuracy of metrics in central venous catheterization without requiring a high performing object detection network by removing false positive predictions identified using uncertainty quantification. The uncertainty for each bounding box was calculated using an entropy equation. The uncertainties were then compared to an uncertainty threshold computed using the optimal point of a Receiver Operating Characteristic curve. Predictions were removed if the uncertainty fell below the predefined threshold. 50 videos were recorded and annotated with ground truth bounding boxes. These bounding boxes were used to train an object detection network, which was used to produce predictive bounding boxes for the test set. This method was evaluated by computing metrics for the predictive bounding boxes with and without having false positives removed and comparing them to ground truth labels using a Pearson Correlation. The Pearson Correlations for the baseline comparisons and the comparisons made using the results calculated using false positive removal were 0.922 and 0.816 for syringe path lengths, 0.753 and 0.510 for ultrasound path lengths, 0.831 and 0.489 for ultrasound usage times, and 0.857 and 0.805 for syringe usage times. This method consistently reduced inflated metrics, making it promising for improving metric accuracy.
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 | 0.007 | 0.041 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".