Accurate and interpretable evaluation of surgical skills from kinematic data using fully convolutional neural networks
This is the companion repository for our paper titled "Accurate and interpretable evaluation of surgical skills from kinematic data using fully convolutional neural networks" published in the International Journal of Computer Assisted Radiology and Surgery - Special Issue of MICCAI 2018, also available on ArXiv.
Architecture
Data
You will need the JIGSAWS dataset to re-run the experiments of the paper.
Requirements
You will need to install the following packages present in the requirements.txt file.
Running the code
Classification
You will first need to run the following: python3 classification.py
.
Then compile the results by running the following: python3 classification.py results
.
Finally, to visualize the class activation map, you will need to run the following: python3 classification.py cas
.
Here is an example of the class activation map for the classification task.
Expert | Novice |
---|---|
![]() |
![]() |
Regression
You will first need to run the following: python3 regression.py
.
Then compile the results by running the following: python3 regression.py results
.
Finally, to visualize the class activation map, you will need to run the following: python3 regression.py cas
.
Here is an example of the class activation map for the regression task.
Suture/needle handling | Quality of the final product |
---|---|
![]() |
![]() |
Reference
If you re-use this work, please cite:
@Article{ismailfawaz2019accurate,
author = {Ismail Fawaz, Hassan and Forestier, Germain and Weber, Jonathan and Idoumghar, Lhassane and Muller, Pierre-Alain},
title = {Accurate and interpretable evaluation of surgical skills from kinematic data using fully convolutional neural networks},
journal = {International Journal of Computer Assisted Radiology and Surgery},
year = {2019}
}