eduzhai > Applied Sciences > Engineering >

Learning and Reasoning with the Graph Structure Representation in Robotic Surgery

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Learning to infer graph representations and performing spatial reasoning in acomplex surgical environment can play a vital role in surgical sceneunderstanding in robotic surgery. For this purpose, we develop an approach togenerate the scene graph and predict surgical interactions between instrumentsand surgical region of interest (ROI) during robot-assisted surgery. We designan attention link function and integrate with a graph parsing network torecognize the surgical interactions. To embed each node with correspondingneighbouring node features, we further incorporate SageConv into the network.The scene graph generation and active edge classification mostly depend on theembedding or feature extraction of node and edge features from complex imagerepresentation. Here, we empirically demonstrate the feature extraction methodsby employing label smoothing weighted loss. Smoothing the hard label can avoidthe over-confident prediction of the model and enhances the featurerepresentation learned by the penultimate layer. To obtain the graph scenelabel, we annotate the bounding box and the instrument-ROI interactions on therobotic scene segmentation challenge 2018 dataset with an experienced clinicalexpert in robotic surgery and employ it to evaluate our propositions.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×