Automatic versus manual tuning of robot-assisted gait training in people with neurological disorders

Introduction: In clinical practice, therapists choose the amount of assistance
that patients receive while walking in a robotic gait trainer. A disadvantage is
that therapists cannot directly feel what the device does. Therefore, algorithms
were developed that automatically adjust the assistance, however,
they have not been compared to the settings that therapists would choose.
Main objective: The goal of this study was to compare the assistance set by
an automatically-tuned (AT) algorithm to manually-tuned (MT) assistance in
a robotic gait trainer.

Methods: Ten participants (6x stroke, 4x spinal cord injury) walked with
both approaches in the LOPES II gait trainer. In both cases, the assistance
was adjusted for various subtasks of walking (e.g. step height). Either the
therapist changed the assistance for each subtask (MT) or the AT algorithm
adjusted the assistance based on errors compared to reference trajectories.

Results and discussion: The different approaches did not always focus on
the same subtasks. On average, participants received less assistance with
the AT algorithm for all subtasks. In spite of this, the largest errors
compared to the reference trajectory were found for the MT approach.
A possible reason for this is that therapists might focus on other factors
while setting the assistance.

Conclusion: An automatically-tuned algorithm can decrease deviations
from a reference trajectory, however, large differences were found
compared to the settings chosen by a therapist and further research should
focus on how this information can be used to optimize robotic gait therapy.