Detection of Suture Needle Using Deep Learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The importance of surgical simulation has increased over the last decade and the majority of medical schools have incorporated simulation into their curriculum. An essential aspect of surgical education is to evaluate how the student performs when compared to an expert surgeon. Another way to evaluate the skill of the student would be by tracking the position of the needle during the procedure, a factor correlating to surgical skill. In this study, we developed deep learning algorithms for needle detection during a video of a surgical procedure. 78 videos of a person doing a running suture on synthetic skin were captured using an HD camera. A total of 3368 images were manually annotated with a VGG annotator tool. Two deep learning algorithms (YOLOv3 and Faster R-CNN) were pretrained on 2219 images extracted from the JIGSAWS dataset, then trained on the 804 images from the training set and finally applied to the 345 images from the evaluation set. The performance of the algorithm was evaluated using the intersection over union (IoU) method as well as by measuring the Euclidean distance between bounding box centroids. These values were compared against the inter-observer reliability among three authors. The best IoU value by deep learning algorithms compared against the ground truth was found to be 0.601 for Faster R-CNN while the average inter-observer value was 0.663. The average Euclidean distances between bounding box centroids for authors and for the Faster R-CNN algorithm were 21.9 pixels and 36.8 pixels, respectively. Through qualitative and quantitative assessment of the algorithm (visually observing the algorithm’s needle annotations), deep learning shows promise for automatically tracking the position of the needle during a suturing operation.
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.
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 | 0.004 | 0.006 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.002 |
| 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 it