JMIR Publications

ABSTRACT

Background:

Lack of functionality is one factor that contributes to high prosthetic rejection rates. Electromyographic (EMG) upper limb prostheses are controlled through muscle contractions in the user’s residual limb, and the incorporation of multi-gesture controls into the Limbitless Solutions upper limb prosthesis requires users to intentionally differentiate between various strengths of muscle contractions to trigger pre-programmed gestures on the device. There is little research on the limitations of expanding device capability. This expansion may lead to a decline in user accuracy and perceived usability, or an increase in training time and cognitive workload.

Objective:

Determine the feasibility of implementing multiple gestures when learning electromyographic controls during a single training session.

Methods:

Scores decreased significantly as zones increased (Kruskal–Wallis H = 21.7, P<.001). The mean (SD) scores were 15.0 (0.0) for one zone, 9.1 (1.1) for three zones, and 5.5 (1.1) for five zones. Perceived usability, as measured by SUS, differed modestly across cohorts (Kruskal–Wallis H₃ = 7.14, P = .068). The progressive cohort achieved the highest average SUS score (mean 76.2 [SD 11.6]), followed by the single-gesture and three-gesture cohorts (each 75.0 [18.9]). The five-gesture cohort rated the system lowest (63.3 [16.2]). Cognitive workload, assessed through NASA-TLX, increased with the number of gestures (Kruskal–Wallis H₃ = 19.9, P< .001). Mean (SD) overall workload scores rose from 23.9 (18.4) for the single-gesture A1 condition to 44.9 (24.7) for the five-gesture A3 condition, with the progressive cohort reporting intermediate workload (37.5 [22.1]). The sample size for quantitative analysis was 54.

Results:

Scores decreased significantly as zones increased (Kruskal–Wallis H = 21.7, P<.001). The mean (SD) scores were 15.0 (0.0) for one zone, 9.1 (1.1) for three zones, and 5.5 (1.1) for five zones. Perceived usability, measured through SUS, differed modestly across cohorts (Kruskal–Wallis H₃ = 7.14, P=.068). The progressive cohort achieved the highest average SUS score (mean 76.2 [SD 11.6]), followed by the single-gesture and three-gesture cohorts (each 75.0 [18.9]). The five-gesture cohort rated the system lowest (63.3 [16.2]). Cognitive workload, assessed through NASA-TLX, increased with the number of gestures (Kruskal–Wallis H₃ = 19.9, P< .001). Mean (SD) overall workload scores rose from 23.9 (18.4) for the single-gesture A1 condition to 44.9 (24.7) for the five-gesture A3 condition, with the progressive cohort reporting intermediate workload (37.5 [22.1]). The sample size for quantitative analysis was 54.

Conclusions:

These findings support the implementation of up to 3 gestures with a progressive mode of training. User preferences trended toward a more complicated system; this may be attributed to perceived improvement. Progressively learning 3 gestures enabled a balance between device capability and retained user intention, perceived usability, and cognitive workload. Clinical Trial: These procedures were approved by the University of Central Florida IRB on October 1, 2024 under STUDY00007067.