Table 1 Microsoft Kinect in Upper Extremity Clinical Applications [49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68]

[14]

Assessment of validity of Kinect v1.0 against marker-based motion capture; 48 normal subjects; upper and lower extremity

Similar reproducibility; different ROM detection for the lower extremity but similar results for shoulder abduction (±3°) and elbow flexion (±11°)

([45, 46]; [47, 48])

Assessment of validity of Kinect v2 for postural control and balance against marker-based motion capture; 30 normal subjects;

High reliability and concurrent validity for balance assessment (trunk, upper and lower extremity kinematics)

[15]

Direct comparison of Kinect against Vicon ® clinical motion capture

Kinect detection is accurate, one order of magnitude less precise than Vicon

[16]

Kinect vs. Vicon for gross and fine movements (controlled study of Parkinson’s disease); movements included whole-body coordinated movements and shoulder flexion/abduction targeted movements

Kinect is highly accurate for gross movement detection, less for smaller hand movements; repeatable measurements (r > 0.9); high interclass correlation for gross extremity/body movements; low correlation for fine hand movements

[19]

Shoulder-specific validity and reliability of Kinect; 10 normal subjects; shoulder joint (flexion, abduction, rotation) assessed in static poses with Kinect, marker based motion analysis, and goniometer; the Kinect was tested both in anterior and sagittal view with insignificant difference in ICC

High reliability, but LOA greater than ±5°, up to 7° for shoulder abduction; Kinect shoulder measurement is most accurate in flexion (high ICC with valid measurements), and least accurate at abduction approaching 90°; note that the analysis focused on extents of motion, not the entire range of motion

[20]

Shoulder ROM by Kinect vs. goniometry; 15 normal and 12 with adhesive capsulitis of the shoulder; Active ROM compared between standard goniometry and Kinect

High ICC; Kinect is repeatable for shoulder ROM measurements (ICCs: 0.91 flexion, 0.94 abduction; 0.91 external rotation); Kinect accurately measures 3D shoulder ROM

[21]

Test-retest repeatability of Kinect for UE, both 12 healthy and 18 stroke subjects; focus on shoulder and elbow kinematics, and spatiotemporal metrics

Study showed acceptable repeatability and sensitivity in both populations; Shoulder and elbow angle measurements all showed greater than 0.9 ICC, indicating repeatability

[17]

Accuracy and reliability of Kinect v2 for clinical measurements – compared with Vicon; 19 normal subjects; spatial range of motion of arm movements evaluated

Most parameters ICC > 0.7; no systematic bias; all joints of the UE and torso detected by Kinect had Pearson correlation > 0.9 against Vicon; concurrent Kinect and Vicon used

[18]

Kinect (anterior) vs. Vicon; 20 normal subjects; balance and arm sway; Kinect and Vicon data collected separately, analyzed for variance in movement patterns and marker positions

Study found that broad movements of the upper extremities had > 90% accuracy, finer hand movements lower accuracy; activities are standardized (game-directed) for comparison between the systems