Automated calibration of somatosensory stimulation using reinforcement learning.
Luigi Borda, Noemi Gozzi, Greta Preatoni and 2 others
PMID 37752607WHAT IT FOUND
The AI tool matched an expert on mapping time and sensation quality, but used fewer stimulations and was faster than the inexperienced and brute-force methods.
It was tested in a small group of healthy and neuropathic people, so it is not proof of patient benefit.
Key findings
01In healthy participants, the AI algorithm mapped nerves in 4.6 ± 2.7 min, similar to the expert's 7.3 ± 3.0 min, and faster than the inexperienced and brute-force methods.
02The AI algorithm delivered fewer stimulations than the expert, averaging 7.1 ± 4.0 compared with 46.2 ± 13.9, while achieving similar sensation quality.
03In neuropathic participants, the platform completed mapping in 6 ± 2 min, used 10.3 ± 3.8 stimulations, and reached a sensation quality index of 0.94 ± 0.01 on a 0 to 1 scale.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
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 online test included only a small number of healthy volunteers and a very small number of neuropathic people. The neuropathic testing was done on a single day, so it does not show whether recalibration would work over time in neuropathy. The study measured calibration speed, number of stimulations, charge, and a custom sensation quality score, not walking, pain, or daily function. Electrode placement still had to be done by the experimenter, so the tool does not make the whole procedure independent. The AI was trained mostly on healthy data, and the authors say more neuropathic participants are needed. Several comparisons with the expert were not statistically significant, so the apparent advantages over an expert are trends.
Declared interests
The supplied text lists funding from Horizon Europe European Research Council, Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, Innosuisse - Schweizerische Agentur für Innovationsförderung, and Swiss Federal Institute of Technology Zurich. No separate author conflict-of-interest declaration is provided.
The easy way to misread this
Do not read this as evidence that TENS or automated calibration improves patient outcomes. The study measured how fast an AI found stimulation settings and how many stimulations it used, not whether patients walked better, had less pain, or gained useful sensation.