# SuFIA: Language-Guided Augmented Dexterity for Robotic Surgical Assistants

**Authors:** Masoud Moghani, Lars Doorenbos, William Chung-Ho Panitch, Sean Huver, Mahdi Azizian, Ken Goldberg, Animesh Garg  
**arXiv:** [2405.05226](https://arxiv.org/abs/2405.05226)  
**Date:** 2024-05-08  
**Summary page:** https://www.summarizepaper.com/en/paper/2405.05226v1/

## Key points

- SuFIA is a groundbreaking framework for natural language-guided augmented dexterity in robotic surgical assistants.
- Developed by Masoud Moghani, Lars Doorenbos, William Chung-Ho Panitch, Sean Huver, Mahdi Azizian, Ken Goldberg, and Animesh Garg.
- SuFIA leverages large language models (LLMs) and perception modules for high-level planning and precise low-level control during surgical sub-task execution.
- Innovative learning-free approach eliminates the need for in-context examples or predefined motion primitives.
- Human-in-the-loop paradigm allows seamless collaboration between surgeon and robotic assistant.
- SuFIA can swiftly restore control to the surgeon in cases of insufficient information or unexpected errors during critical tasks to ensure patient safety.
- Rigorously evaluated through simulations and physical experiments demonstrating autonomous performance of surgical procedures with precision and efficiency.
- Represents a significant advancement in robotic surgery by enabling enhanced dexterity and autonomy through natural language guidance.

## Summary

SuFIA is a groundbreaking framework designed for natural language-guided augmented dexterity in robotic surgical assistants. Developed by Masoud Moghani, Lars Doorenbos, William Chung-Ho Panitch, Sean Huver, Mahdi Azizian, Ken Goldberg, and Animesh Garg, SuFIA leverages the advanced reasoning capabilities of large language models (LLMs) along with perception modules to enable high-level planning and precise low-level control of robotic systems during surgical sub-task execution. What sets SuFIA apart is its innovative learning-free approach that eliminates the need for in-context examples or predefined motion primitives. One of the key features of SuFIA is its human-in-the-loop paradigm, which allows for seamless collaboration between the surgeon and the robotic assistant. In cases where there is insufficient information or unexpected errors arise during mission-critical tasks, SuFIA can swiftly restore control to the surgeon to ensure patient safety and optimal outcomes. The effectiveness of SuFIA has been rigorously evaluated through simulations involving four surgical sub-tasks and physical experiments on a surgical robotic platform in a lab setting. These tests have demonstrated SuFIA's ability to autonomously perform common surgical procedures with precision and efficiency even under challenging physical and workspace conditions. Overall, SuFIA represents a significant advancement in the field of robotic surgery by enabling enhanced dexterity and autonomy through natural language guidance. Its successful implementation holds promise for improving surgical outcomes and enhancing the capabilities of robotic surgical assistants in real-world clinical settings. For more information about SuFIA, visit their project website at orbit-surgical.github.io/sufia.