Autonomous Robotic Assistant for Knee-Replacement
Background and Motivation
Paramjit Baweja, Shivangi Gupta, Li-Wei Yang, Qilin Wu, Abhishek Warrier
What needs to be done
Building a robotic system that can assist in Knee Replacement Surgery. We are focusing on the fundamental aspect of the surgery – drilling holes.
We would like to express our gratitude to our advisors Kaushik Balasundar, Parker Hill, Sundaram Seivur, and Dr. Branko Jaramaz from Smith+Nephew for their invaluable guidance.
We also extend our thanks to our mentors from the MRSD program, Professors John Dolan and Dimi Apostolopoulos, for their support and encouragement.
What we want to replace
Acknowledgements
Perception Pipeline
The System
Step 1 : Annotate points on the femur and tibia and use Meta’s Segment-Anything to extract masks of just the bone.
Step 2 : Use Meta’s Co-Tracker model to track the points to enable re-registration and motion compensation.
Step 3 : Register the bone models using ICP + Custom Heuristics.
System Architecture
ç
ç
Results
ç
Left : A hole drilled into the tibia with perfect accuracy
Bottom: Surgical Pins inserted into adjacent holes, showing orientation accuracy
KUKA LBR Med 7
DREMEL
Our desired position accuracy (in both static bones and surgery with motion) was < 2 mm
Top: A laser pointer showing our end effector falling within desired accuracy
Right: The drilled hole has sub 2mm accuracy
D405
E-STOP