Journal of Rehabilitation Medicine 51-6 | Page 64

458 M. van Bloemendaal et al. 360 cm (schematic representation in Fig. S1 1 ). The camera was used both in a stationary position (a length of 130 cm of the walkway could be captured reliably within the field of view) and as a moving camera on a 7-m long rail (dolly track) placed parallel to the walkway, over which the camera could be moved manually (Fig. 1). An overview of the SGAS requirements and costs is presented in Appendix S1 1 . The SGAS measures calibrated positions on the floor of the walkway. The calibration procedure is based on the method proposed by Zhang (31). Intrinsic camera parameters are determined from approximately 30 images of a planar 6×10 chequerboard pattern of 9-cm squares (Fig. 2). These para- meters characterize the camera’s optical system. Placing this chequerboard pattern vertically in a well-defined location on the walkway sets up an orthogonal laboratory coordinate system in which the y-axis runs along the walkway, the x-axis is perpendi- cular to the walkway, and the z-axis points vertically upwards. The position and orientation of the camera with respect to this coordinate system are determined from a single image of the chequerboard pattern at this location. This image provides the information for the camera’s extrinsic parameters. Combining the intrinsic and extrinsic parameters results in the camera’s pro- jection matrix, which describes how the coordinates of a point in the laboratory coordinate system are converted into the pixel coordinates of the camera’s image plane. In the current study, the set-up and calibration process of the SGAS took approximately 10 min; calibration was repeated after assessing 8 subjects and in a clinical setting requires one calibration for the day. The y-coordinates of an object on the floor are determined manually from the video image. On the basis of the projection matrix, the SGAS software draws a thin red line in the video image representing the projection of the line in the x-direction for a given y-coordinate in the plane z = 0 (Fig. 3). The user adjusts the y-coordinate of this line by moving the computer’s mouse until the projection matches the position of the object on the floor in the image; for instance, the location of heel contact of a foot. The user reads the corresponding y-coordinate from the SGAS user interface. The time of the event is derived from the video frame rate (time resolution 0.02 s). Counting the number of frames yields the time difference between 2 events in the video recording. Simultaneously with the SGAS, the gait of the subject was recorded with the GAITRite® system. The GAITRite® system (GAITRite® Platinum 488P, CIR Systems Inc., New York, USA) consists of a portable carpet walkway embedded with pressure-activated sensors that sample at 60 Hz. The walkway is 488 cm long and 61 cm wide and contains an active sensor area of 384 × 48 sensors arranged 1.27 cm from each other (centre on centre, 18,432 sensors in total). Experimental set-up During a 30-min test session, subjects were tested in 4 different walking conditions in a fixed order under stationary camera set- up: barefoot walking at comfortable speed; barefoot walking at slow speed; barefoot toe walking at comfortable speed; and shod walking at comfortable speed wearing their own comfortable flat-soled shoes. Data collection with the SGAS and GAITRite® system was conducted by one investigator (MVB). Ten valid gait trials, 5 in which a left and 5 in which a right footstep was visible within the 130-cm field of view, per walking condition were collected with the GAITRite® system and SGAS in a stationary position. Gait strides were collected in the given field of view by the stationary SGAS camera for the conditions of slow speed and toe walking. In addition to the stationary camera conditions, while walking barefoot at a comfortable walking speed, 4 gait trials, including 4–8 strides per trial, were collected with the SGAS camera being moved along the walkway by the investigator. Inter-rater reliability was assessed with 3 trained observers who were instructed in the definitions of the spatiotemporal gait parameters (Table SI 1 ) and gait analysis method. To assess intra- rater reliability, one observer (MVB) assessed the same data on 2 different occasions (minimally 1 month apart). Inter-rater and intra-rater reliability were assessed for the barefoot comfortable walking condition with the stationary SGAS camera. Data processing and analysis The initial contact (heel or toe) and toe-off distance and time- points during each trial were identified manually from the video images of the SGAS by 5 trained observers and were recorded in a Microsoft Excel spreadsheet format (Microsoft Corporation, Washington, USA) that was designed to automatically calculate the spatiotemporal gait parameters (https://github.com/Mvan- Bloemendaal/SGAS). Analysed data from the SGAS and the GAITRite® system were: step length, step time, stance time, double support time, stride length, stride time, and swing time. http://www.medicaljournals.se/jrm/content/?doi=10.2340/16501977-2559 1 Fig. 3. Video analysis using the spatiotemporal gait analysis system (SGAS). www.medicaljournals.se/jrm