Based on the linked source; automatically prepared and checked against the original report.
Study design
The pilot study involved eight adults with no prior manual wheelchair experience. Researchers used an ABA single-subject design involving six experimental participants and two controls. Baseline and post-test assessments took place in a real wheelchair using a SMARTWheel, while intervention sessions used the gamified miWe virtual reality simulator. The simulator included a convolutional neural network classifier to assess propulsion patterns.
Propulsion results
Experimental participants increased their use of the semicircular propulsion pattern, which clinical guidelines recommend. Tau-U values ranged from 0.33 to 0.67, with 100% non-overlapping data across the experimental group. The reported results indicated consistent improvement during the intervention and retention at post-test. Control participants instead showed deterioration, with negative Tau-U values and no non-overlapping data.
Other measures and limitations
Contact angle also improved among experimental participants, reaching a Tau-U of 0.67 and 100% non-overlapping data, but did not improve in controls. Cadence results varied, with Tau-U values from -0.67 to 0 and mixed non-overlapping-data results, providing no consistent evidence of reduction after testing. The findings support further evaluation of automated visual feedback in virtual-reality wheelchair training, but this record summarizes a pilot abstract rather than the full paper.
