University of Twente Student Theses
Artificial intelligence for assessing operation time and surgical difficulty in laparoscopic cholecystectomy
Oosterhoff, V.P.S. (2024) Artificial intelligence for assessing operation time and surgical difficulty in laparoscopic cholecystectomy.
|
PDF
8MB |
| Abstract: | The operating room (OR) is responsible for significant revenue and costs in hospitals. Therefore, carefully scheduling the operating time is of great importance. However, surgeries frequently exceed their planned duration due to unforeseen intraoperative challenges. To address this issue and enhance surgical scheduling, we proposed a method that utilizes both preoperative and intraoperative variables to estimate operating time dynamically. Our studies at Meander Medical Center (MMC) focused on laparoscopic cholecystectomy, a high-volume procedure characterized by substantial intraoperative variability. Our initial analysis examined surgical data from 2017 to 2023 to quantify the extent of operating time inaccuracies. Machine learning (ML) models were trained on available clinical data to predict operating time. A linear regression (LR) model achieved a root mean squared error (RMSE) of $\pm$ 14.18 minutes, marginally outperforming the conventional method’s RMSE of $\pm$ 16.22 minutes. Despite this improvement, the predictive accuracy remains insufficient for clinical application. In a subsequent prospective analysis, additional patient-specific factors were incorporated to enhance predictive performance. Following data cleaning, 199 patients were retained from an initial cohort of 231. The best-performing LR model achieved an RMSE of $\pm$ 12.28 minutes, representing a modest improvement over existing methods but still falling short of the precision required for clinical implementation. To further advance the model, intraoperative variables were integrated through an assessment of surgical difficulty using the Nassar scale. The multi-scale vision transformer (MViTv2) was employed to classify surgical videos, with the best model achieving an accuracy of 36$\%$. This level of performance remains inadequate for clinical use. While these studies have contributed to a deeper understanding of the factors influencing operating time, the predictive models developed thus far are not yet suitable for clinical deployment. Nonetheless, further refinement and integration of preoperative and intraoperative variables hold promise for more accurate operating time predictions in the future. |
| Item Type: | Essay (Master) |
| Faculty: | TNW: Science and Technology |
| Subject: | 44 medicine |
| Programme: | Technical Medicine MSc (60033) |
| Link to this item: | https://purl.utwente.nl/essays/104464 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page