Predicting Anastomosis or Stump Leakage After Laparoscopic Gastrectomy: A Deep Learning Approach to Intraoperative Image Analysis
Bibliographic record
Abstract
PURPOSE: Postoperative leakage is a critical complication of laparoscopic gastrectomy for gastric cancer. Predicting leakage during surgery can enhance patient outcomes by enabling a timely intervention. This study aimed to develop and validate deep learning models for predicting leakage using laparoscopic images of the anastomosis sites. MATERIALS AND METHODS: We analyzed 10,256 laparoscopic images from 2,035 patients who underwent gastrectomy for gastric cancer at three institutions. Six datasets (EXP1 to EXP6) were created based on variations in image quality and analytical methods. Six deep learning architectures, ResNet18, ResNet34, ResNet50, EfficientNet_V2_L, Inception_V3, and DenseNet121, were employed for training. Deep learning models were trained to classify images into normal or leakage categories at the duodenal stump (DS) and esophagojejunal (EJ) anastomoses. Model performance was evaluated using F1 scores, recall, and Grad-CAM visualization. RESULTS: Leakage was identified in 1.3% and 4.3% of the patients with DS and EJ, respectively. Among the six datasets, EXP1, which used one image per patient and applied augmentation, exhibited the best performance. ResNet18 trained on EXP1 demonstrates the highest recall values, achieving 0.8474 for DS and 0.8000 for EJ, with F1 scores of 0.6357 and 0.6938, respectively. Grad-CAM revealed that both local and surrounding tissue features were critical for model prediction. CONCLUSIONS: Deep learning could predict leakage during gastric cancer surgery. High-resolution imaging, single-image analysis, and data augmentation were pivotal for model performance. These findings lay the groundwork for clinical applications and future research on surgical image analysis.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.004 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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".