danikiyasseh / SAIS

[Nature Biomedical Engineering 2023] Decoding surgical activity from videos with a vision transformer

Home Page:https://www.nature.com/articles/s41551-023-01010-8

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

SAIS

License: CC BY-NC-SA 4.0

SAIS is a surgical AI system that decodes surgical activity based exclusively on surgical videos. This system was first introduced in a 2023 Nature Biomedical Engineering paper.

Features

SAIS has the ability to decode the following elements of surgery:

  • surgical steps (e.g., dissection vs. suturing)
  • suturing sub-phases (e.g., needle handling vs. needle driving vs. needle withdrawal)
  • four distinct suturing gestures
  • six distinct tissue dissection gestures
  • needle handling binary skill level
  • needle driving binary skill level

Inference

To demonstrate the utility of SAIS, we have developed an interface that allows users to upload a brief surgical video (on the order of 10-20 seconds) and identify:

  • six distinct tissue dissection gestures
  • binary skill level of suturing activity

About

[Nature Biomedical Engineering 2023] Decoding surgical activity from videos with a vision transformer

https://www.nature.com/articles/s41551-023-01010-8


Languages

Language:Python 100.0%