Automatic phase prediction from low-level surgical activities

This is the companion web page for our paper titled "Automatic phase prediction from low-level surgical activities".

This paper has been accepted for publication to the International Conference on Information Processing in Computer-Assisted Interventions IPCAI 2015.

If you have any comments/questions regarding the research work or software, please feel free to contact us.

Example of a decision-tree

An example of a decision-tree and its corresponding set of rules able to predict surgical phases are available here: decision-tree-ipcai2015.txt.

The software

The software is developed in Java, it uses the Weka classification library.

The package is licensed under a GPLv3 (license in the Jar files) : donwload the soure code.

The package contains a sample file (sample.arff) that can be used to test the program.

To run the program just execute: java -jar ipcai2015.jar sample.arff in a console.

Ouput of the method on the sample data

#Single activity:
Number of leaves: 116
Size of the tree: 128

Correctly Classified Instances        2488               89.2396 %
Incorrectly Classified Instances       300               10.7604 %
Kappa statistic                          0.8238
Mean absolute error                      0.0552
Root mean squared error                  0.1661
Relative absolute error                 22.8094 %
Root relative squared error             47.7729 %
Total Number of Instances             2788     

Correctly Classified Instances : 0.9813486370157819

#With clustering:
#Cluster 1 Single Activity :
Number of leaves: 30
Size of the tree: 33

Correctly Classified Instances         677               86.6837 %
Incorrectly Classified Instances       104               13.3163 %
Kappa statistic                          0.8063
Mean absolute error                      0.0666
Root mean squared error                  0.1825
Relative absolute error                 24.801  %
Root relative squared error             49.831  %
Total Number of Instances              781     

#Cluster 2 Single Activity :
Number of leaves: 97
Size of the tree: 108

Correctly Classified Instances        1810               90.1844 %
Incorrectly Classified Instances       197                9.8156 %
Kappa statistic                          0.8126
Mean absolute error                      0.0517
Root mean squared error                  0.1607
Relative absolute error                 24.3687 %
Root relative squared error             49.396  %
Total Number of Instances             2007     

#Cluster 1 Local Context :
Correctly Classified Instances : 0.9871959026888605

#Cluster 2 Local Context :
Correctly Classified Instances : 0.9785749875435974

Last update : April 2015