search

SlicerIGT / OMTCentralLineTraining

Inexpensive open-source system for practicing hand-hand coordination for US guided needle insertions.
1 stars 0 forks source link

Hardware Dependencies #1

Open oroszgabor83HU opened 2 years ago

oroszgabor83HU commented 2 years ago

Hi Mark, please help me with this issue. If I have a - let's say - "non-PLUS compatible" US machine (not listed in the PLUS manual sources) how can I technically build up a training system similar to yours? Is it any restriction (e.g. brands that cannot be used)? Where can I find some source for this?

Thanks in advance, Gabor

ungi commented 2 years ago

If your ultrasound machine has any standard video output (e.g. for external monitor), then you could use a frame grabber to feed the ultrasound video to PLUS. Currently, Epiphan and Imaging Controls frame grabbers are supported by PLUS. Note that this is a bit outdated project. We now try to move towards using video camera (webcam) only for computerized evaluation of simulated medical procedures. Position trackers are always hard to mount on tools, especially lightweight things like needles. But AI video processing can recognize all tools e.g. for central line insertion. We can provide better feedback now from a single camera than what we could do earlier with trackers. Accurate aiming with the needle is of course important, but there are indirect signs on an external video feed that can verify accurate needle placement, e.g. colored fluid in the syringe.

oroszgabor83HU commented 2 years ago

Hi Tamas, thanks for the fast answer! Sounds very interesting the video tracking/processing with AI. Do You have any public, available source for this project? I am an intensivist guy from Hungary and would be interested in this issue in the aspect of other invasive procedure (e.g. percutaneous tracheostomy).

Thanks, Gabor

ungi commented 2 years ago

Our source code is in this repository: https://github.com/SlicerIGT/aigt Under DeepLearnLive. It is in an early experimental phase, not a solid foundation to build on yet. I would just start collecting videos of training procedures. If the quality is good for a human rater to evaluate, then sooner or later it will be good enough for AI as well.

oroszgabor83HU commented 2 years ago

Thank You very much! G