Stereoscopic scene flow for robotic assisted minimally invasive surgery

Area of Science:

• Medical imaging
• Computer vision
• Surgical robotics

Background:

• Accurate 3D shape and motion data of surgical sites are crucial for image-guidance and robotic control in minimally invasive surgery.
• Current methods may lack the precision or efficiency needed for real-time surgical applications.

Purpose of the Study:

• To develop and validate a scene flow algorithm for reconstructing dense 3D structure and deformation from stereoscopic surgical videos.
• To improve metric measurements for enhanced image-guidance and robotic control during minimally invasive procedures.

Main Methods:

• A scene flow algorithm that propagates information from seed matches to reconstruct 3D structure and deformation.
• Imposition of spatial and temporal constraints for accurate and efficient dense 3D scene flow reconstruction.
• Validation using simulated data with varying noise levels and benchmark phantom model data.

Main Results:

• The algorithm accurately and efficiently reconstructs dense 3D scene flow.
• Validation demonstrated robustness against image noise in simulation.
• Successful qualitative results were obtained from in vivo robotic-assisted surgical videos.

Conclusions:

• The proposed scene flow algorithm provides accurate 3D reconstruction of surgical site dynamics.
• This method has practical value for enhancing image-guidance and robotic control in minimally invasive surgery.
• The approach shows promise for real-time applications in surgical robotics.