Force-feedback sensory substitution using supervised recurrent learning for robotic-assisted surgery.

Area of Science:

• Robotics
• Surgical Technology
• Biomedical Engineering

Background:

• Force feedback is a critical limitation in robot-assisted minimally invasive surgery.
• Add-on sensors are impractical for clinical settings, necessitating alternative solutions.

Purpose of the Study:

• To develop and validate a non-invasive force estimation approach for robot-assisted surgery.
• To improve the accuracy and stability of force sensing during surgical procedures.

Main Methods:

• Reconstruction of a 3D tissue surface deformation structure by minimizing an energy functional.
• Application of a Recurrent Neural Network-Long Short Term Memory (RNN-LSTM) architecture for force estimation.

Main Results:

• The proposed method demonstrates long-term stability in force estimation.
• Significant accuracy improvements, ranging from 54% to 78%, were achieved compared to existing methods.

Conclusions:

• The developed force estimation approach effectively addresses the lack of force feedback in robot-assisted surgery.
• This non-invasive technique offers a practical and accurate solution for enhancing surgical performance and safety.