[ARTICLE] Design and test of an automated version of the modified Jebsen test of hand function using Microsoft Kinect – Full Text



The present paper describes the design and evaluation of an automated version of the Modified Jebsen Test of Hand Function (MJT) based on the Microsoft Kinect sensor.


The MJT was administered twice to 11 chronic stroke subjects with varying degrees of hand function deficits. The test times of the MJT were evaluated manually by a therapist using a stopwatch, and automatically using the Microsoft Kinect sensor. The ground truth times were assessed based on inspection of the video-recordings. The agreement between the methods was evaluated along with the test-retest performance.


The results from Bland-Altman analysis showed better agreement between the ground truth times and the automatic MJT time evaluations compared to the agreement between the ground truth times and the times estimated by the therapist. The results from the test-retest performance showed that the subjects significantly improved their performance in several subtests of the MJT, indicating a practice effect.


The results from the test showed that the Kinect can be used for automating the MJT.


Deficits in motor function, in the form of hemiparesis or hemiplegia, are a frequent consequence of cerebral stroke [1]. Even though motor function may be regained to some extent through intensive rehabilitative training following acute treatment of stroke, deficits in hand function often remain [23]. Following discharge from the rehabilitation unit, patients are typically asked to perform unsupervised self-training in their own home. The lack of supervision during training at home will likely have an impact on the patient’s training compliance and training quality. Therefore, it is important to perform regular evaluations of the patient’s functional level in order to provide useful supervision and to maintain patient motivation. The patients’ performance in a specific motor function test provides valuable insight into whether the training scheme chosen for a patient is effective or it should be changed. Thus, it is very important that the motor function tests being used are objective and reflect the actual functional level of the patient being tested. Several validated motor function tests including assessment of hand function exist, e.g. Jebsen Test of Hand Function [4], Action Research Arm Test [5], Fugl-Meyer Assessment [6], Wolf Motor Function Test (WMFT) [7], Box and Blocks Test [8] and Nine Hole Peg Test [9]. Common for all these tests is that they must be administered by a therapist, which might be a source for variability in the test results, and cause the test results not always to be completely reproducible and objective. In tests including performance time as an outcome measure, e.g. the WMFT, the reaction time of the subject could introduce a bias to the results, as suggested by previous studies [1011]. Likewise, the end time of the test would likely be subjected to a bias, since the examiner has a finite reaction time. Thus, both the reaction time of the examiner and the subject could be potential sources of bias and variability in timed motor function tests. The sensitivity of a motor function test is affected by sources of bias and variability and therefore it is of interest to minimize these, to make detection of even small changes possible.

By automating motor function tests, the objectivity of the tests would be increased. This might also make possible to use the tests at remote sites, without direct supervision, as a part of a tele-rehabilitation service. Finally, automated tests could be administered more frequently. Previous studies have shown that selected parts of the WMFT can be automated by use of motion sensors mounted on the body of healthy subjects [10] and stroke patients [11]. Both systems automated the test by analyzing three-dimensional kinematics data from body-worn sensors (inertial measurement units) mounted on the most affected wrist, arm and shoulder of stroke patients [1011]. Similarly, using inertial measurement unit sensors, Yang et al. (2013) showed that when administering the 10 m walking test, the output from their system was in close agreement with the walking speeds estimated using a stop-watch [12]. These systems require though correct positioning and mounting of the motion sensors [10]. Huang et al. (2012) showed that also a computer vision based approach, consisting of a monitor camera and a Xilinx Virtex II Pro Field Programmable Gate Array (for computation), may be used for automating the WMFT. All participants being tested had to wear a black sweatband on the wrist of the extremity being tested [13]. Another low-price method for capturing the movements of a patient performing a motor function test is the Microsoft Kinect sensor (Kinect). By using a Kinect, the need for body mounted sensors is eliminated, thus lowering the susceptibility to data loss and easing donning and doffing of the system. Furthermore, the Microsoft Kinect sensor is a low-cost commercially available device. In this paper, we describe the design and test of a Kinect based system for automatic evaluation of a standardized, validated motor function test, administered to stroke patients with hand function deficits. The Modified Jebsen Test of Hand Function (MJT) [14], initially proposed by Bovend’Eerdt et al. (2004) as a test for assessment of gross functional dexterity in stroke patients, was selected for automation as this test is easy to administer and takes short time to complete.

Continue —> Design and test of an automated version of the modified Jebsen test of hand function using Microsoft Kinect | Journal of NeuroEngineering and Rehabilitation | Full Text

Fig. 3 The edge of the table was detected in the binary image (lower) produced by thresholding the depth image (upper) into two parts, one part containing all pixels with a depth value lower than a depth level of 300 mm below the surface of the table and the other part containing pixels with depth values above this threshold


, , , , , , , , , , , ,

  1. Leave a comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: