Enhancing motion tracking accuracy of a low-cost 3D video sensor using a biomechanical model, sensor fusion, and deep learning.
Shahar Agami, Raziel Riemer, Sigal Berman
PMID 36188943WHAT IT FOUND
Fusing a Kinect camera with a goniometer made wrist position and elbow angle estimates closer to reference markers in eight healthy people.
The tracker also improved these estimates, but patients were not tested.
Key findings
01Offline fusion of Kinect and goniometer data produced smaller wrist position errors (34.6 mm) and elbow angle errors (0.1°) than direct Kinect measurement (75.6 mm and 18.9°).
02Run-time LSTM trackers had lower wrist position errors (42.5 mm with goniometer input, 40.1 mm without goniometer input) than direct Kinect measurement (75.6 mm).
03The network without goniometer input had a lower elbow angle error (5.5°) than the network with goniometer input (7.1°).
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The tracking methods were tested only in eight healthy participants, not in people with stroke. The study did not test clinical rehabilitation outcomes or patient recovery. The setup used a fixed chair, restricted trunk motion, and a manipulator that immobilised the wrist and elbow rotation, so it does not show how the method works in more natural reaching. The offline optimization required a goniometer attached to the arm, which is less convenient than marker-less tracking alone. The tested model assumed reaching or pointing with little change in arm orientation and did not address elbow supination-pronation.
The easy way to misread this
Do not read the accuracy numbers as evidence that this improves stroke rehabilitation outcomes. It was tested in eight healthy participants doing reaching tasks, not in stroke survivors, and it did not measure patient recovery.