Towards Retrieving Force Feedback in Robotic-Assisted Surgery: A Supervised Neuro-Recurrent-Vision Approach
IEEE Transactions on Haptics
|January 24, 2017
Summary
This study introduces a novel vision-based approach using deep learning to estimate forces in robotic surgery, enhancing surgeon precision by restoring the sense of touch without physical sensors.
Area of Science:
- Robotics
- Computer Vision
- Surgical Technology
Background:
- Robotic-assisted minimally invasive surgeries offer advantages but lack crucial force feedback.
- This limitation hinders surgeon's tactile sense, potentially impacting procedural precision.
Purpose of the Study:
- To develop a novel force estimation method for robotic surgery.
- To provide surgeons with tactile feedback by estimating applied forces using visual data.
Main Methods:
- A vision-based solution combined with supervised learning (LSTM-RNN) for force estimation.
- Extracting heart surface geometry via energy functional minimization to recover 3D structure.
- Training a deep network to map visual-geometric information to applied force.
Main Results:
- The approach successfully estimates applied forces without physical sensors.
- Evaluated on phantoms and realistic tissues, achieving an average root-mean-square error of 0.02 N.
- Demonstrates a viable alternative to traditional force sensing devices.
Conclusions:
- The proposed vision-based deep learning method effectively estimates forces in robotic surgery.
- This technique overcomes limitations of physical force sensors, improving safety and precision.
- Enables enhanced tactile feedback for surgeons, advancing minimally invasive procedures.
