A Novel, Wearable Inertial Measurement Unit for Stroke Survivors: Validity, Acceptability, and Usability

Abstract

Objective: To establish the concurrent validity, acceptability, and sensor optimization of a consumer-grade, wearable, multi-sensor system to capture quantity and quality metrics of mobility and upper limb movements in stroke survivors.

Design: Single-session, cross-sectional.

Setting: Clinical research laboratory.

Participants: Thirty chronic stroke survivors (age 57 (10) years; 33% female) with mild to severe motor impairments participated.

Interventions: Not Applicable.

Main outcome measures: Participants donned 5 sensors and performed standardized assessments of mobility and upper limb (UL) movement. True/false, positive/negative time in active movement for the UL were calculated and compared to criterion-standards using an accuracy rate. Bland-Altman plots and linear regression models were used to establish concurrent validity of UL movement counts, step counts, and stance time symmetry of MiGo against established criterion-standard measures. Acceptability and sensor optimization were assessed through an end-user survey and decision matrix.

Results: Mobility metrics showed excellent association with criterion-standards for step counts (video: r=0.988, P<.001, IMU: r=0.921, P<.001) and stance-time symmetry (r=0.722, P<.001). In the UL, movement counts showed excellent to good agreement (paretic: r=0.849, P<.001, nonparetic: r=0.672, P<.001). Accuracy of active movement time was 85.2% (paretic) and 88.0% (nonparetic) UL. Most participants (63.3%) had difficulty donning/doffing the sensors. Acceptability was high (4.2/5).

Conclusions: The sensors demonstrated excellent concurrent validity for mobility metrics and UL movements of stroke survivors. Acceptability of the system was high, but alternative wristbands should be considered.

Keywords: Gait; IMU; Stroke; Upper limb; Wearables.

Fig 1

(A) The MiGo system consists of 5 sensors placed at bilateral wrists, paretic hip and bilateral ankles. (B) Picture of the wristwatch. (C) collected from caregivers of study participants and stroke researchers and Picture of the clip placed at the hip and on each ankle.

Fig 2

Bland–Altman plots demonstrating limits of agreement between MiGo and Video annotated movement count for the paretic and nonparetic UL. Dashed lines represent upper and lower limits of agreement.

Fig 3

(A) Bland-Altman plot demonstrating limits of agreement between MiGo and Video annotation. (B) Agreement between MiGo and video annotation (*r = 0.988, p < 0.001) (C) Bland-Altman plot illustrating limits of agreement between MiGo and APDM. (D) Agreement between MiGo and APDM (*r = 0.921, p < 0.001).

Fig 4

Stance time symmetry agreement between MiGo and APDM for each FAC level. The bolded black line represents the overall agreement between MiGo and APDM with the corresponding dashed lines and symbols representing each FAC level, respectively.

Fig 5

(A) Bar plots showing perceived comfort, appearance, and willingness to wear each sensor daily. (B) Bar plots showing mean time to don and doff each sensor. *Neurotypical data represents data is included only for reference.