11. カテーテルの性能評価
Objective evaluation of
catheters/interventional skills
Delft University of Technology
80% correct
Fig. 4 Distribution of subject groups in the PCA space.
Decision boundary 70.9% correct
Fig. 3 Distribution of AFVs for two catheters in the PCA space.
Objective Evaluation Approach of Catheters and
Interventional Manipulation Skills
Aki Kunikoshi, Helene Clogenson, Jenny Dankelman
Delft University of Technology, Mechanical Maritime and Materials Engineering (3mE)
Mekelweg 2, 2628 CD Delft, The Netherlands
email: a.kunikoshi@tudelft.nl
Minimally Invasive Surgery and
Interventional Techniques
Integrated Interventional Imaging
Operating System
The IIIOS project has received funding from the European Community s Seventh Framework Programme (FP7/2007-2013) under Grant Agreement no 238802
Delft University of Technology
Current evaluation techniques
Human raters are required
Only based on elapsed time to accomplish tasks
An objective approach to evaluate
interventional manipulation skills
Purpose
Methods Result
Catheter prediction
Subject group prediction
Automatic catheter activity detection
Activity Feature Vector (AFV)
Sample 55 videos of the task
Subjects: 7 novice residents
Catheters: Renal Double Curve (RDC) and Cobra 2 (C2)
Task: To Manipulate a catheter so that it passes one
bifurcation in a phantom
Fig. 1 A model of typical vascular geometry and dimensions.
(twaiting, tCT_rot., tCT_ret., tCT_ret., tGW_adv., tGW_ret.)
AFVs were submitted to principal component analysis
(PCA) and the 1st-3rd principal components were used to
perform linear discriminant analysis (LDA).
Fig. 2 Flowchart of how to classify catheter activities.
Performance of the proposed method was evaluated using
a leave-one-out cross-validation (55 validations in total).
Subjects can be grouped into four groups depends on
their location on LDA plane.
Possibility to classify residents based on their skills?