Rowe, Justin B., et al. "Robotic assistance for training finger movement using a Hebbian model: a randomized controlled trial." Neurorehabilitation and neural repair 31.8 (2017): 769-780.

January 10, 2022

This is another interesting work from David Reinkensmeyer's group on understanding the role of robotic assistance. He is probably the only person who has pursued this topic consistently in the last 20 years.

Robot-assisted therapy has been found to be effective and not inferior to conventional therapy, although the mechanism for recovery remains unknown. There are several possible mechanisms through which robot-assisted therapy could be beneficial:

1. The intensity of therapy by allowing patients to train more movements with assistance than is possible without assistance.
2. Provision of proprioceptive feedback time-synchronized with movement intention which can leverage Hebbian-type or spike time-dependent plasticity type mechanisms to remodel circuit to bring about behavioral changes.
3. Motivation results from the ability to perform movements with robotic assistance.

The current evidence in favor of robotic assistance in driving recovery is equivocal. There are studies that suggest a positive role for robotic assistance, and studies that do not show any added benefit of robotic assistance. The authors also claim that some of these studies had controlled for the number of movements performed by the different groups undergoing therapy.

There are also negative effects of robotic assistance such as “slacking” which can reduce a patient's effort during training, and also diminish error in their movements. These can be detrimental to motor learning/recovery. The authors also talk about effort, but the precise definition of effort is not given. It is possible that different studies use slightly different definitions.

The goal of this study was to compare the effects of high versus low levels of assistance with the FINGER robot while playing Guitar Hero. The primary outcome was the BBT score, and a set of secondary measures were also used: FMA, NHPT, IMI, MAL, robot-assisted proprioceptive assessment, self-actuation measure, and Geriatric depression score,

Therapy protocol:

• Two groups: High assistance (HA) vs. Low assistance (LA). The assistance was controlled to ensure that the HA and LA groups had a consistent success rate of 80% and 60% in the game. The controller used a compliant position controller that did not diminish errors. A previous study was carried out to demonstrate that the controller does not lead to slacking.
• Chronic stroke patients with a single stroke, and with a minimal score on BBT were recruited.