Nonlinear Regression of Remaining Surgical Duration via Bayesian LSTM-Based Deep Negative Correlation Learning

Wu, Junyang; Tao, Rong; Zheng, Guoyan

doi:10.1007/978-3-031-16449-1_40

Cited by 1 publication

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed TransLocal model yields the best performance with MAE = 7.1 min for all surgeries, MAE = 10.6 for long surgeries and MAE = 4.4 for short surgeries. Compared to other SOTA techniques that employ the same annotation, 17 , 19 TransLocal provides a slight improvement in terms of MAE. However, in this study, we employed a smaller dataset (80 vs. 120 videos) and the performance was tested on more folds (10 vs. 4).…”

Section: Discussionmentioning

confidence: 92%

“…The other models show worse performance but this difference was not proved statistically significant except for LSTM and CNN‐LSTM ( p > 0.05). In addition, the table shows the results for two of the best techniques in, 17 which were evaluated on 120 operations (Cholec120), and for a recent model 19 evaluated on Cholec80. The TransLocal pipeline presents better performance than the SOTA methods despite being trained on fewer surgeries or/and tested on more splits.…”

Section: Resultsmentioning

confidence: 99%

“…In a recent study, Wu et al. introduced a Bayesian decorrelated network (BD‐Net) that trains an ensemble of Bayesian LSTMs using deep negative correlation learning 19 . In the proposed implementation there are two branches for RSD (main task) and phase/surgeon experience prediction (auxiliary tasks).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Prediction of remaining surgery duration in laparoscopic videos based on visual saliency and the transformer network

Loukas,

Seimenis,

Prevezanou

et al. 2024

Robotics Computer Surgery

View full text Add to dashboard Cite

BackgroundReal‐time prediction of the remaining surgery duration (RSD) is important for optimal scheduling of resources in the operating room.MethodsWe focus on the intraoperative prediction of RSD from laparoscopic video. An extensive evaluation of seven common deep learning models, a proposed one based on the Transformer architecture (TransLocal) and four baseline approaches, is presented. The proposed pipeline includes a CNN‐LSTM for feature extraction from salient regions within short video segments and a Transformer with local attention mechanisms.ResultsUsing the Cholec80 dataset, TransLocal yielded the best performance (mean absolute error (MAE) = 7.1 min). For long and short surgeries, the MAE was 10.6 and 4.4 min, respectively. Thirty minutes before the end of surgery MAE = 6.2 min, 7.2 and 5.5 min for all long and short surgeries, respectively.ConclusionsThe proposed technique achieves state‐of‐the‐art results. In the future, we aim to incorporate intraoperative indicators and pre‐operative data.

show abstract

Section: Discussionmentioning

confidence: 92%

Section: Resultsmentioning

confidence: 99%