Design and participants
A cross-sectional, repeated measure, analytical study was designed to examine the intra-individual reliability and concurrent criterion-related validity of the smartphone iPhone 4® in the ETGUG test. Five participants were recruited from a healthy over-65 population. The exclusion criteria were a history of pain in the last twelve months, or a history of surgery, malignancy, or musculoskeletal disorders of any limb that might be aggravated by the procedures involved in the test. Informed consent was obtained from all subjects, and study procedures were consistent with the Helsinki declaration. The study was approved by the local University Committee.
Inertial sensors
The participants wore the two sensors overlapping in a small neoprene sleeve, placed at the level of the middle third of the sternum. Previous studies on the variability of inertial sensor measurements placed at the levels of different body segments [11, 12] show that sensors located at the level of the sternum provide reliable data. The first of the sensors used was an Inertiacube3® (IC3; Intersense, Bedford, MA, USA). This module integrates two two-axes accelerometer, three singles-axis gyroscopes and a three-axis magnetometer compass within low volume (26.2 × 39.2 × 14.8 mm3). InertiaCube3 combines the aforementioned sensing elements with an integrated Kalman filtering algorithm. The unit can provide orientation and gravity compensated acceleration information aligned with the Earth’s magnetic north. InertiaCube3 can measured accelerations up to ±6 g [13], in this study were studied only the values obtained from the accelerometer subunit.
The second sensor (the one to be validated in the present study) was that incorporated in the iPhone 4® of Apple®. This smartphone is equipped, as is the IC3, with three triaxial elements for the detection of kinematic variables: a gyroscope, a magnetometer, and an accelerometer. Apple uses an LIS302DL accelerometer in the iPhone 4® [14]. Kinematic data were acquired using the xSensor® Pro software of Crossbow Technology, Inc. This app couldn’t record at higher sampling rates than 32 Hz. A previous study [7] had validated the iPhone 4®’s gyroscope, showing the inter-observer error to be 4.0° (standard deviation of the difference between measurements).
The orientation and movement of these two sensors are presented as Euler angles RPY (roll, pitch, and yaw). If the sensor’s RPY axes are aligned with the anatomical axes of the trunk, the roll angle of a movement is around the anteroposterior (AP) axis [15]. Movements in this plane are less frequent than those in other planes [16]. The pitch angle is around the left-right (or mediolateral, ML) axis [15], and the yaw angle is around the vertical (V) axis [15].
Test protocol
The subjects performed the ETGUG test three times. They used a chair without armrests, and were instructed orally not to use their arms to stand up or sit down. Various studies have explored this test using chairs without armrests [17–19]. This choice could reduce inter-subject variability by eliminating the option of whether or not to use the arms in the standing up and sitting down phases. The ETGUG test selected was that using a 10 metres long corridor, the aim being to include as many gait cycles during the test as possible [19]. The beginning and end of the ten metres were marked on the floor with 2½-cm wide tape. The protocol was as follows:
-
1)
The subject sat with his or her back in contact with the back of the chair (the seat was 460 mm high and lacks armrests).
-
2)
The ETGUG begins with the therapist’s go sign and the subject stands up (Sit-to-Stand).
-
3)
The subject begins walking ten meters (Gait Go).
-
4)
The subject turns around a wide tape mark placed 10 m away from the chair (Turn).
-
5)
The subject walks back toward the chair (Gait Come).
-
6)
The subject turns away from the chair to sit down (Turn-to-Stand-to-Sit).
To evaluate the test-retest reliability of the iPhone 4®’s kinematic measurements, the participants repeated the protocol three times. The first and second trials were completed consecutively for reliability and after the two sensors were removed from the participant’s body following an hour of rest, the sensors were put back in the same position, and the ETGUG test protocol was repeated to evaluate the stability of the measurement after removing and then putting back the two sensors. It was assumed that the participant’s performance remained unchanged within the time of this resting period. The same examiner used the same device and the same protocol to place the devices, and to conduct the ETGUG test.
Sub-phases of the expanded timed-get-up-and-go test
After the test, and based on the analysis of the accelerometry data of all the participants, the ETGUG test was divided into 5 main sub-stages: from sitting to standing (Sit-to-Stand, Si-St), gait going away (Gait Go, GG), turn (T), gait coming back (Gait Come, GC), and turning to sit down (Turn-to-Stand-to-Sit, T-St-Si) (Figure 1). The different sub-phases were detected with the sensor parameters as follows:
-
For the identification and analysis of the sitting-standing transitions (Si-St and T-St-Si), we followed a previously published protocol [20] (Figure 1).
-
For the identification and analysis of the turn transition (T), we also followed a previously published method [21] in which 180° rotations are detected by analysing the yaw rotation signal, which should identify the first turn at ten metres and the second turn which is made in order to sit down and thus terminate the test (Figure 1).
-
The identification and analysis of the GG and GC sub-phases was performed by analysing the data remaining once the Si-St and T-St-Si [20] and T [21] sub-phases had been detected and delimited.
Signal processing
The IC3 data were acquired at a 100 Hz sampling frequency, and those of the iPhone 4® at 32 Hz. Data processing was performed off-line, expanding and synchronizing the two sets of time series using the basic package of the R® software environment.
Statistical analysis
We performed descriptive statistics with measures of central tendency and dispersion of the maximum and minimum peak acceleration in AP, ML, V and the acceleration magnitude (AM). The AM is a vector that has the same effect on the system as the component vectors. AM is calculated from the three acceleration vectors of the axes of motion (x, y, z) as the square root of the sum of the values of the three axes (AM = √x2 + y2 + z2). Analyses of the relations between the scores obtained by simultaneously from IC3 and iPhone 4® were performed. The relations were studied by using the Coefficient of Multiple Correlation (CMC) [22]. Cohen and Holiday criteria [23] were applied to interpret these correlation coefficients, suggesting the following categorization: very low correlation for values below 0.20; low correlation for values between 0.20 and 0.39; moderate correlation for values between 0.40 and 0.69; high correlation for values between 0.70 and 0.89 and very high correlation for values above 0.89. To evaluate the reliability of the iPhone 4® measurements, we calculated the intraclass correlation coefficient (ICC) between the two sensors acceleration signal. Values between 0.70 and 0.95 were considered acceptable reliability indicators [23]. To investigate statistical agreement between the two sensors we used the Bland-Altman method [24].
