University of Twente Student Theses


AI-based prediction model of surgical difficulty in laparoscopic cholecystectomy

Egging, R.M. (2021) AI-based prediction model of surgical difficulty in laparoscopic cholecystectomy.

[img] PDF
Abstract:Laparoscopic cholecystectomy (LC) is the standard procedure to remove the gallbladder. Although this procedure has evolved to a relatively safe and tolerable daycare procedure, it can be difficult at times and complications can arise. Complicated gallstone disease, such as cholecystitis or gallstone pancreatitis, are risk factors for increased technical difficulty of a LC. Although it is possible to make a preoperative prediction of the surgical difficulty, perioperative findings can be surprising. Understanding the difficulty of the surgical scenario with AI-based models is important to allow benchmarking in surgical performance and improve planning on the OR. This study aimed to develop a Deep Learning (DL) to predict the difficulty of laparoscopic cholecystectomy on specific operative findings. A difficulty grading scale was used, based on the Nassar score. To train the DL network, frames were extracted from the recorded videos. All frames were labeled for ‘gallbladder’ difficulty grade 1-3 and ‘adhesions’ difficulty grade 1-3. Frames consisting of out-of-body images or in which the gallbladder was not visible were excluded. This resulted in a total of 26.483 frames. A ResNet was used as a backbone for the model. Hyperparameters were tuned to improve model results. Both multiclass and binary classification networks were trained. The network that was trained to classify gallbladder difficulty (3-grades) performed better (accuracy 74%) than the network trained to classify adhesions difficulty. It is possible to classify cholecystitis with an accuracy of 91% and classify easy cases with an accuracy of 87%. The results of this study could be used as a starting point for further research in classifying difficulty in LC. This is a first step to improve understanding of surgical scenery and allow benchmarking for surgeons in LC.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page