217

Journal of Vestibular Research 23 (2013) 217–225 DOI 10.3233/VES-130489 IOS Press

Ambulatory balance monitoring using a wireless attachable three-axis accelerometer Soo-Chan Kima,1 , Mi Joo Kimb,1 , Nambeom Kimc , Jong Hyun Hwangd and Gyu Cheol Hand,∗ a

Deptartment of Electrical and Electronic Engineering, Hankyong National University, Anseong, Korea Departments of Otorhinolaryngology Head and Neck Surgery, College of Medicine, Yonsei University, Seoul, Korea c Neuroscience Research Institute, Graduate School of Medicine, Gachon University, Incheon, Korea d Deptartment of Otolaryngology Head and Neck Surgery, Graduate School of Medicine, Gachon University of Medicine and Science, Incheon, Korea b

Received 17 August 2012 Accepted 11 July 2013

Abstract. BACKGROUND AND OBJECTIVES: The ability of conventional diagnostic equipment to monitor feelings of dizziness experienced during daily activities is limited. Our goal is to develop an ambulatory multipurpose device for monitoring balance to prevent falling in daily life. MATERIALS AND METHODS: A three-axis accelerometers and gyroscope sensors were attached to the head, pelvis, and legs of vestibular neuritis (VN) patients or age-, height-, and body weight-matched healthy volunteers. The sum of the deviations for the scalar value of acceleration [signal vector magnitude, SVM (g)] and angular velocity (◦ /s) was measured using the modified Romberg test. RESULTS: The repeated measure ANOVA model with acceleration showed a greater group difference (p < 0.001) than that with angular velocity (p < 0.01). There was no significant interaction effect within-subjects factor between replication and groups (p < 0.178). SVM within the VN group significantly increased for all sensor locations compared to the control group (p < 0.01). Strong correlations between measurements taken at head and pelvis as sensor location were observed for both groups (VN/control, r = 0.68/r = 072). CONCLUSION: The SVM appears to accurately assess balance while standing, even repetitive measurement or any location in body. Keywords: Dizziness, acceleration, balance, monitoring, falling

1. Introduction Vertigo results from a tonic imbalance between the vestibular systems [1]. To evaluate equilibrium, physicians run various clinical tests such as positioning/positional tests, oculomotor tests, caloric tests, 1 These

authors contributed equally to this work. author: Gyu Cheol Han, Department of Otolaryngology-Head and Neck Surgery, Gachon University of Medicine and Science, Graduate School of Medicine, 1198 Guwoldong, Namdong-gu, Incheon, 405-760, Korea. Tel.: +82 32 460 3324; Fax: +82 32 467 9044; E-mail: [email protected]. ∗ Corresponding

rotation chair tests, and dynamic posturography [2]. These methods do not sufficiently reflect vertigo experienced in daily life except in the context of an active disease state, and do not provide information about falling [3]. Moreover, those medical equipments are not only ambulatory, but also unsuitable for screening for dizziness [4]. In those dizziness cases with test results that are normal, primary physicians face difficulties with making an accurate diagnosis and deciding appropriate treatment [5]. The repeated evaluation of balance is generally obligated to monitor the vestibular compensation process in vertiginous patients for proper rehabilitation [6–8].

c 2013 – IOS Press and the authors. All rights reserved ISSN 0957-4271/13/$27.50 

218

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

Patients with acute vestibular loss sway their body to avoid falling [4]. This reflexive response leads to postural imbalance. If compensatory body perturbation is not adequate, the COG is moved excessively and the individual falls towards the side with the lesion [4]. This physical manifestation of a compensatory vestibulo-spinal response can be represented by the Romberg test [9]. However, it is very challenging to record or analyze the COG because compensatory body perturbation is extremely complicated and human body is not a true inverted pendulum when performing daily activities [10]. To overcome these obstacles, various sensors have been used such as ones that measure images, strain gauge, inclination, inertia, acceleration, and angular velocity. Acceleration (g, cm/s2 ) and angular velocity (◦ /s) sensors are often implemented the most because these devices are relatively easy to build in compact forms, generate a large amount of data, and consume low quantities of energy. Acceleration sensors provide information on the dynamics of fast movements and inclination (linear translation). On the other hand, angular velocity sensors monitor the angular dynamics (rotation) around a rotational axis [11–13]. Our goal is to develop an ambulatory multipurpose device for long-term monitoring of balance during daily activities. We also desired to incorporate a function to warn against the possibility of falling, and have this device evolve into a 24-hour Holter-like monitoring system in the future. To develop an ideal device suitable for our purpose, we assembled a tiny accelerometer and gyroscope sensor with memory circuits and a small battery system in one case so that the device could be attached wherever on the human body. To manufacture an ideal balance-monitoring device, several preconditions should be satisfied or must be identified including mechanical completeness and clinical applicability. The device we designed was initially used to perform a modified Romberg test to estimate the algorithm performance. We compared acceleration or angular velocity under the same conditions. We also evaluated the learning effect, and identified the most suitable sensor location on the body.

Table 1 Device specifications Power Acceleration range Angular velocity range Duration of continuous use Sampling rate Multichannel capability

2.7 ∼ 3.3 V ± 16G (± 8 G) ± 2000◦ /s 48 h (CR2 battery) 50 Hz No limitation

(GIRBA2248). We conducted a case-control study with healthy volunteers (y = 43.53 ± 15.32; n = 18; male/female ratio = 4/14) and vestibular neuritis (VN) patients (y = 45.65 ± 12.18; n = 17; male/female ratio = 10/7, Right /Left side lesion = 10/7). We noted the age (p = 0.85), height (cm, VN/control = 164.1 ± 6.52/160.7 ± 5.8, p = 0.99), weight (kg, VN/control = 65.4 ± 9.2/58.9 ± 7.2, p = 0.77), and body mass index (BMI) of all participants. VN was diagnosed according to the Coates criteria [14]. Recruited patients referred to a tertiary referral hospital within 3 d of the onset of their symptoms without any anti-vertiginous medication. The value of canal paresis and spontaneous nystagmus in VN group was 41.98 ± 11.61% (range from 28.06 to 77.69), and 12.23 ± 9.20◦/s (range from 3 to 38), respectively. The modified Romberg test was done 2 d after the laboratory tests due to patient’s safety. Healthy volunteers were defined by absence of neurootologic diseases. 2.2. Modified Romberg test

2. Methods

To minimize proprioception, we used 60 × 60 × 30 cm high-resilience foam (TaeKang sponge, Seoul, South Korea). The subjects were placed on the center of the foam where the shape of their feet had been drawn. They were instructed to hold their hands 15 cm in front of their chest and maintain an upright position. Three separate trials were done with the eyes closed while standing on the foam for 15 s. We analyzed the averaging value of three trials. If the subjects fell or stepped down during the experiment, these sessions were disregarded and repeated from the beginning. All tests were performed in the location of the facility with constant luminance, temperature, and humidity, and with guardians present to ensure the safety of the patients.

2.1. Subjects

2.3. Sensor system

All clinical experiments were carried out from August 2011 to April 2012 with the approval of the Gachon University Institutional Review Board

We assembled a three-axis accelerometer (ADXL 345; Analog Device, Norwood, MA, USA) and a threeaxis gyroscope sensor (ITG3200; InvenSense, Sunny-

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

219

Fig. 1. The study device (left) main board included a three-axis accelerometer and a gyroscope sensor with a microcontroller (right) assembled sensor module with a battery and secure-digital memory in a case.

vale, CA, USA) in a case. The sensor module has six degrees of freedom. Whole memory circuit and rechargeable battery systems were packaged with the sensors in one case (Table 1, Fig. 1). Data generated by each sensor were sent to a microcontroller (Atmega 328; Atmel, San Jose, CA, USA) via I2C communication and stored in integrated secure-digital memory. We calibrated the scale and offset of the digital sensors with respect to acceleration and angular velocity estimated using commercial inertia sensors (MotionNode; GLI Interactive, Seattle, WA, USA) according to the manufacturer’s specifications [15,16]. We based the time-axis synchronization between sensor modules on simultaneous changes of angular velocity during data acquisition. 2.4. Sensors location To accommodate continuous changes in the center of gravity (COG) and to record physical momentum for posture regulation, the sensors were placed on the head, pelvis, and both lower legs. The head was selected because it is near to both vestibular systems. The pelvis was also selected because it is close to the center of gravity (COG), and both legs were selected to monitor body sway in lower level as posturogrphy, and to record falling. A sensor was attached to the center of the forehead with a hair band. Another sensor was attached to the middle of the pelvis or the front lower abdomen connecting both sides of the anterior superior iliac spine, and two sensors were attached to the lateral side of each leg using Velcro and elastic bandages.

All sensors were placed in the same spatial alignment. Acceleration and angular velocity data include vector information and provided the same directional information across sensors and subjects. The sensors were set to the y-axis as the front, x-axis as vertical to the ground and opposite to the direction of gravity, and z-axis as the right side of the subject (Fig. 2). 2.5. Data processing Acceleration and angular velocity data were saved in DOS format. All analyses were performed with LabVIEW (National Instruments, Austin, TX, USA). The sampling rate was 50 Hz and the window period for each trial was 15 s. Quantitative analyses including frequency domain and time domain analyses, or acceleration and angular velocity analyses were then performed. 2.6. Signal vector magnitude Accelerometers and gyroscopes have distinct characteristics and functions. An accelerometer can sense vibrations and has the ability to gauge tilt relative to the Earth’s surface, but cannot measure rotation. A gyroscope has the capability of measuring the rate of rotation (angular velocity) around a particular axis. It helps measure orientation using the principles of angular momentum whereas an accelerometer measures linear motion (Fig. 2). Gyroscopes depend on the rotational axis when measuring angular velocity [17]. In other words, sensitivity for measuring angular velocity

220

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

Fig. 2. Three-axis directionality of the sensors and sensor locations. This figure illustrates the relationship between linear acceleration (ah and ap ) and angular velocity (a and a ). All sensors were arranged in the same direction on the forehead, pelvis between both anterior superior iliac spine, and both legs. The ah represent the a , and the ap represent the a under the assumption that the radius was sufficient (r). ah : acceleration on head; a : angular velocity on head; ap : acceleration on pelvis; a : angular velocity on pelvis; r: radius; θ: sway angle.

is not high when the rotational axis is altered. Therefore, recording angular velocity in a highly accurate manner is very challenging because body motion is very irregular while performing daily activities. From another point of view, body sway at the central axis of the sensor itself can be interpreted as horizontal and vertical movement rather than rotation. Therefore, the linear physical momentum at a single defined point of circular movement under the assumption that the radius was sufficient was calculated with the following formulae (1 ∼ 4): a∼ = a = dθ dt l= r·θ ω=

2

(rω) v2 = = rω 2 r r

(1) (2)

The three axes (x, y, and z) could be compared separately using body sway as an index, but we used the scalar values of the axes for analysis. The changing amount of acceleration or angular velocity is zero when there is no movement, so the measured values can be used to quantify the dynamics of balance control. Because angular velocity is not affected by gravity while acceleration is always influenced by gravity, we had to consider the offset value of each selected section to account for the constant effect of gravity. Thus, we calculated the standard deviations of the three-axis signals from each acceleration and angular velocity sensor. We used Eq. (5) to determine the degree of body sway along the three axes [18,19]: SVM =

 x 2 + y 2 + z2

(5)

(3)

dl dθ =r· = r·ω ∼ (4) =v dt dt For this case, angular momentum can be interpreted as linear motion when an object is in uniform circular motion with a radius of “r”. In the formulae above, l represents the circumference of the rotational circle, ω is the angular velocity, υ is the linear velocity at the circumference, a represents the linear acceleration at the circumference, and a is angular acceleration at the circumference (Fig. 2).

In this formula, x, y, and z are the standard deviations of the angular velocity and acceleration values along each axis. The square of each value represents the variances along each axis. Equation (5) provided a numerical value representing the degree of body sway [defined as the signal vector magnitude, SVM, (g)] using the summation of variances along the three axes. This SVM had a very small scale because this variable represented the dynamics of maintaining a stationary state that are very diminutive.

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

2.7. Statistical analysis We checked the normality of the data with a Kolmogorov-Smirnova test before performing parametric tests such as repeated measure (RM) ANOVA and Pearson’s correlation. No variable was rejected, indicating that they followed a normal distribution. The data are presented in a box – whisker plot. To compare the mean difference of the mean SVM values (mean value of three replication) between the healthy control and VN group, we preformed independent two-sample t-test following by Levene’s test and generated 95% confidence intervals. We used a two-way RM ANOVA to identify significant differences among data from the three replicate experiments and four sensor locations for each subject, and Greenhouse-Geisser correction to adjust the violation of sphericity of the covariance. In addition, intraclass correlation coefficient (ICC) was reported to support the reliability of repeated measurements. Pearson’s correlation was used to test the significance of the summed scalar values for each sensor location. All statistical analysis was performed using SPSS v17.0 (SPSS Inc., Chicago, IL, USA), P-value less than 0.05 was considered statistically significant.

3. Results 3.1. SVM versus angular velocity We compared SVM and angular velocity as measurements of subtle movements made by the subjects attempting to remain stationary. We performed RM ANOVA model fitting using SVM and angular velocity as explanatory variables. F–values for betweensubjects effects provided RM ANOVA model with SVM had bigger group difference [F(1,33) = 14.8, p < 0.001] than that with angular velocity [F(1,33) = 7.4, p < 0.01]. We therefore decided to use SVM as an explanatory variable. ICCs of RM ANOVA model using SVM were 0.53 and 0.52 for normal and VN, respectively. 3.2. Learning effect on the triplicate trials There was no significant replication effect [F(1.176, 38.795) = 1.879, p < 0.178] without interaction effect between Replication and Group [Replication*Group, F(1.176,38.795) = 1.716, p < 0.199]. That suggested that the mean value was not different for each replication within each group, implying that there was no learning effect during replication measurement (Table 2).

221

Table 2 RM ANOVA results Tests for between-subjects effects Source df Mean square (× 10−5 ) F Sig. Group 1 5.72 14.81 0.001 Error 33 0.39 Tests for within-subjects effects Source df Mean square (× 10−5 ) Locations 1.69 3.91 Locations * Group 1.69 2.76 Error (Locations) 55.76 0.51 Replication 1.18 2.88 Replication * Group 1.18 2.63 Error (Replication) 38.79 1.53

F Sig. 7.69 0.002 5.43 0.01 1.88 0.178 1.72 0.199

Sig: the significance level; df: corrected df with GreenhouseGeisser’s method, *interaction.

3.3. Effects of sensor location To compare the SVM at each location between the control and VN groups, we used mean value of the three replications because there was no significant difference among replication measurements in the withinsubjects effect analysis of RM ANOVA. We found significant interaction effect between Location and Group in the SVM dynamics of the head, pelvis, and legs [Location *Group, F(1.69,55.76) = 5.429, p < 0.01] (Fig. 3). Independent two-sample t-test revealed the head sensors showed the most prominent mean difference between the control and VN group (head: mean difference = 0.003658, t(16.2) = 4.738, p < 0.001) (Table 3 and Fig. 3). In addition, the pelvis and both leg measurements also showed significant mean differences between the two groups (pelvis: mean difference = 0.0020, t(16.3) = 2.97, p < 0.009, right leg: mean difference = 0.0025, t(16.2) = 2.9, p < 0.005, left leg: mean difference 0.0020, t(16.3) = 3.3, p < 0.004). 3.4. Relation between the legs and the other parts To identify the strength of the inter-positional associations among the modified Romberg measurements taken at different locations such as the head and pelvis, we calculated the Pearson’s correlation coefficient for each pair-wise combination of locations. Strong correlations between measurements taken at different locations were observed for both groups. Additionally, correlations between the pelvis and right leg locations (VN/control, r = 0.91/0.92) were the greatest for both groups. The weakest correlations were found between the head and right leg locations (VN/control, r = 0.65/0.63). The measurements from the head and left leg (Fig. 4) showed the greatest difference in correlations between the groups (VN/control, r = 0.81/0.60).

222

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer Table 3 Descriptive statistics

Sensor location

Group

Head

Normal VN∗ Normal VN∗ Normal VN∗ Normal VN∗

Pelvis Right Leg Left Leg ∗ VN

Minimum (× 10−4 ) 2.33 9.11 1.88 3.81 1.83 6.97 1.83 2.56

Maximum (× 10−4 ) 12.18 111.95 13.05 119.41 11.47 160.55 13.29 106.86

Median (× 10−4 ) 4.90 32.16 3.61 14.09 3.05 19.81 3.55 16.86

IQR (× 10−4 ) 4.56 50.81 2.22 18.56 1.76 20.01 2.43 25.58

Mean (× 10−4 ) 5.96 42.54 4.63 25.09 3.88 29.10 4.40 24.44

Std. error (× 10−4 ) 0.70 7.92 0.71 6.84 0.61 8.62 0.64 6.03

P-value < 0.001 < 0.01 < 0.01 < 0.01

means vestibuloneuritis; std error means standard error.

Fig. 3. Box and whisker plot showing the Romberg measurements according to study group and sensor location. The signal vector magnitude was explained detail in the text. Results from the head sensors showed the most prominent mean difference [mean difference = 0.003658, t(16.2) = 4.738, p < 0.001]. The pelvis and both leg measurements also showed significant mean differences between the two groups [body: mean difference = 0.0020, t(16.3) = 2.97, p < 0.009, right leg: mean difference = 0.0025, t(33) = 3.0, p < 0.005, left leg: mean difference 0.0020, t(16.3) = 3.3, p < 0.004]. *VN means vestibuloneuritis. ∗∗ p < 0.001. ◦: 0.75/0.25 percentile ± 1.5 x IRQ, + : 0.75/0.25 percentile ± 3x IRQ.

4. Discussion According to previous studies, we ascertained that acceleration of head, pelvis, and even both legs could accurately assess balance in standing position. A traditional Romberg test is performed by asking the subject to stand at attention with the feet together and eyes open. The subjects are then asked to

close their eyes for 15 s. Falling or body sway is observed and the test administrator determines the subject’s balance [20]. In this study, we modified traditional Romberg test for subjects to perform it with greater difficulty by adding the same examination, but performed on foam. We measured the SVM of body movement for 15 s at each trial. We assumed that shorter testing time might have helped patients to fo-

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

223

Fig. 4. Correlations between modified Romberg test scores taken at different locations. Strong correlations between measurements taken at different locations were found for both groups. The strongest correlation was observed between the mid-body (pelvis) and right leg locations (VN/control, r = 0.91/0.92) for the two groups. The weakest correlation was between the head and right leg locations (VN/control, r = 0.65/0.63). Measurements for the head and left leg showed the greatest differences in correlations between the two groups (VN/control, r = 0.81/0.60). ∗ VN means vestibuloneuritis.

cus more on the test. In practice, the testing time of Romberg test has been various from 6 to 60 s according to purpose of the study [9,21,22]. Another reason we modified the testing time was for the seniors since they may not be able to fully participate in the test due to muscular fatigue. In addition, reduced testing time may improve the inter-test and inter-individual reliability. Furthermore, many subjects taking the Romberg test would be expected to suffer from acute dizziness, making it difficult for them to maintain their balance for a long period of time. The Romberg test is performed to diagnose sensory ataxia [23]. It has been used in clinics for over 150 y [20,24], and its ability to objectively evaluate the relationship between human bipedal locomotion and the vestibular system has been verified several times [23,25,26]. However, the Romberg test is not quantitative and has several limitations such as low diagnostic sensitivity and specificity, low discrimination power for lesion side or level, and weak prediction of falling risk [21]. Inexpensive force transducers on a fixed support surface have been used to the Romberg

test to provide an indirect measurement of body sway while standing [10,27]. Nevertheless, this type of medical equipment is insufficient for evaluating a rebalance strategy of the body during high frequency motion stimulation [28–30]. Other implemented devices with the same purpose are too heavy to use while performing daily activities and it could not check multiple points of the body simultaneously. Aside from age [22,31,32], other proven factors that affect Romberg test results are height and weight [33]. We calculate Pearson’s coefficient of correlation between BMI and modified Romberg measurements taken at each sensor location. Modified Romberg measurements taken at the head were correlated with BMI (r = 0.61). Modified Romberg measurements obtained at the other body locations showed a weak correlation with BMI. This finding indicates that the head is a BMI-sensitive location compared to the others locations. This could be a considerable factor when interpreting the study results. The SVM could be affected by height than angular velocity, because the radius of gyration in head is bigger than that is in the

224

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer

pelvis or legs. However, everybody has different COG point, and the purpose of this study is trace the COG sways, SVM is directly proportional to radius of gyration. Therefore, we consider the value of SVM/BMI or SVM/ height as sway index especially in head (Fig. 2). At the moment of swaying, the human body operated reflective strategy to prevent falling. The representatives were ankle and hip strategy, and paraspinal reflex [4,10,25,31]. The strategy selection depends on disease status and age [31,34]. The ankle strategy is the fastest response especially to an event of postural imbalance [34]. Therefore, we compared SVM measurement on the head and pelvis with the legs. We, from the comparison of SVM measurement, have decided to position the sensors on the pelvis and head where body sway is well reflected. Although our statistical analysis revealed a significant difference of SVM between the groups, we did not control for age or gender differences. Additionally, some data points deviated significantly from the average although most body sway movements were very subtle, indicating that the study subjects made sporadic larger movements. These factors should be considered when further refining the modified Romberg test. We could not detect any learning effect to the repeated modified Romberg test trials performed in our study. This might be attributed to test duration. Therefore, comparisons of data according to testing time should be performed to more decisively evaluate the learning effect. Our goal is to develop an ambulatory multipurpose device for long-term monitoring of body balance to prevent falling in daily life. Our device could also be used for gait analysis because of quiet small volume. In order to decrease the discomfort of the subjects, caused by wires that were attached to multiple sensors, we designed devices capable of wireless communication for sending data through radio signals that were received centrally to a single device. Otherwise, the data was stored in each device. Our design has the advantage of scalability to multiple channels because adding more channels would not affect the frequency of data transmission or incur the risk of data loss [35,36]. Therefore, a consecutive application study should be performed in the future.

5. Conclusions Results of the present study demonstrated the possibility of designing an algorithm and portable device for

monitoring balance while standing. In summary, this device would be used to monitor balance during daily activities in the future.

Acknowledgements This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (NRF-20100011858).

Conflict of interest The authors declare no competing financial interests. The authors alone are responsible for the content and writing of the paper.

References [1]

M. Dieterich, Central vestibular disorders, J Neurol 254 (2007), 559–568. [2] I.S. Curthoys, The interpretation of clinical tests of peripheral vestibular function, Laryngoscope 122 (2012), 1342–1352. [3] J.P. Staab, Chronic subjective dizziness. Continuum (Minneap Minn), 18 (2012), 1118–1141. [4] J.H. Allum, M.G. Carpenter and A.L. Adkin, Balance control analysis as a method for screening and identifying balance deficits, Ann N Y Acad Sci 942 (2001), 413–427. [5] E.E. Lang and R. McConn Walsh, Vestibular function testing, Ir J Med Sci 179 (2010), 173–178. [6] F.B. Horak, D.M. Wrisley and J. Frank, The Balance Evaluation Systems Test (BESTest) to differentiate balance deficits, Phys Ther 89 (2009), 484–498. [7] J.M. Furman, Role of posturography in the management of vestibular patients, Otolaryngol Head Neck Surg 112 (1995), 8–15. [8] J.M. Furman, Posturography: uses and limitations, Baillieres Clin Neurol 3 (1994), 501–513. [9] A. Khasnis and R.M. Gokula, Romberg’s test, J Postgrad Med 49 (2003), 169–172. [10] J.H. Allum and A.L. Adkin, Improvements in trunk sway observed for stance and gait tasks during recovery from an acute unilateral peripheral vestibular deficit, Audiol Neurootol 8 (2003), 286–302. [11] A. Olivares, G. Olivares, F. Mula, J.M. Gorriz and J. Ramirez, Wagyromag: Wireless sensor network for monitoring and processing human body movement in healthcare applications, Journal of Systems Architecture 57 (2011), 905–915. [12] K.J. O’Donovan, R. Kamnik, D.T. O’Keeffe and G.M. Lyons, An inertial and magnetic sensor based technique for joint angle measurement, Journal of Biomechanics 40 (2007), 2604– 2611. [13] A.M. Sabatini, C. Martelloni, S. Scapellato and F. Cavallo, Assessment of walking features from foot inertial sensing. Biomedical Engineering, IEEE Transactions on 52 (2005), 486–494.

S.C. Kim et al. / Ambulatory balance monitoring using a wireless attachable three-axis accelerometer [14]

A. Coates, Vestibular neuronitis, Trans Am Acad Ophthalmol Otolaryngol 73 (1969), 395–408. [15] S. Scapellato, F. Cavallo, C. Martelloni and A.M. Sabatini, Inuse calibration of body-mounted gyroscopes for applications in gait analysis, Sensors and Actuators A: Physical 123-124 (2005), 418–422. [16] F. Ferraris, U. Grimaldi and M. Parvis, Procedure for effortless in-field calibration of three-axis rate gyros and accelerometers, Sensors Mater 7 (1995), 311–330. [17] H.J. Luinge and P.H. Veltink, Measuring orientation of human body segments using miniature gyroscopes and accelerometers, Medical and Biological Engineering and Computing 43 (2005), 273–282. [18] D.M. Karantonis, M.R. Narayanan, M. Mathie, N.H. Lovell and B.G. Celler, Implementation of a real-time human movement classifier using a triaxial accelerometer for ambulatory monitoring. Information Technology in Biomedicine, IEEE Transactions on 10 (2006), 156–167. [19] M.H. Lee, J. Kim, S.H. Jee and S.K. Yoo, Integrated Solution for Physical Activity Monitoring Based on Mobile Phone and PC, Healthc Inform Res 17 (2011), 76–86. [20] D.J. Lanska and C.G. Goetz, Romberg’s sign: development, adoption, and adaptation in the 19th century, Neurology 55 (2000), 1201–1206. [21] K.A. McMichael, J. Vander Bilt, L. Lavery, E. Rodriguez and M. Ganguli, Simple balance and mobility tests can assess falls risk when cognition is impaired, Geriatr Nurs 29 (2008), 311– 323. [22] T.C. Hain, Approach to the patient with dizziness and vertigo. in: Practical Neurology, J. Biller ed., Lipincott Raven Publishers, Philadelphia, 1997, p. 159. [23] H.K. El-Kashlan, N.T. Shepard, A.M. Asher, M. SmithWheelock and S.A. Telian, Evaluation of clinical measures of equilibrium, Laryngoscope 108 (1998), 311–319. [24] N. Reicke, The Romberg head-shake test within the scope of equilibrium diagnosis, HNO 40 (1992), 195–201. [25] C.D. Ford-Smith, J.F. Wyman, R.K.J. Elswick, T. Fernandez and R.A. Newton, Test-retest reliability of the sensory organization test in noninstitutionalized older adults, Arch Phys Med Rehabil 76 (1995), 77–81. [26] F.B. Horak, C. Jones-Rycewicz and F.O. Black, Effects of

225

vestibular rehabilitation on dizziness and imbalance, Otolaryngol Head Neck Surg 106 (1992), 175–180. [27] L.M. Nashner and J.F. Peters, Dynamic posturography in the diagnosis and management of dizziness and balance disorders, Neurol Clin 8 (1990), 331–349. [28] J.H. Allum, B.R. Bloem, M.G. Carpenter and F. Honegger, Differential diagnosis of proprioceptive and vestibular deficits using dynamic support-surface posturography, Gait Posture 14 (2001), 217–226. [29] H. Kingma, G.C. Gauchard, C. de Waele, C. van Nechel, A. Bisdorff, A. Yelnik, M. Magnusson and P.P. Perrin, Stocktaking on the development of posturography for clinical use, J Vestib Res 21 (2011), 117–125. [30] B. Najafi, K. Aminian, F. Loew, Y. Blanc and P.A. Robert, Measurement of stand-sit and sit-stand transitions using a miniature gyroscope and its application in fall risk evaluation in the elderly, IEEE Trans Biomed Eng 49 (2002), 843–851. [31] J. Gill, J.H. Allum, M.G. Carpenter, M. Held-Ziolkowska, A.L. Adkin, F. Honegger and K. Pierchala, Trunk sway measures of postural stability during clinical balance tests: effects of age, J Gerontol A Biol Sci Med Sci 56 (2001), M438–447. [32] B.D. Iverson, M.R. Gossman, S.A. Shaddeau and M.E. Turner, Jr., Balance performance, force production, and activity levels in noninstitutionalized men 60 to 90 years of age, Phys Ther 70 (1990), 348–355. [33] A.A. Fabunmi and C.A. Gbiri, Relationship between balance performance in the elderly and some anthropometric variables, Afr J Med Med Sci 37 (2008), 321–326. [34] J.H. Allum, M.G. Carpenter, F. Honegger, A.L. Adkin and B.R. Bloem, Age-dependent variations in the directional sensitivity of balance corrections and compensatory arm movements in man, J Physiol 542 (2002), 643–663. [35] J.H. Bae, M.J. Kim, E.J. Lee, G.C. Han and S.-C. Kim, Abstracts, A walking test using acceleration sensor and holter as a new vertigo examination. The 24th ARO midwinter research meeting, 2011, p. 71, poster #516. [36] M.J. Kim, T.W. Kim, S.J. Jang, A.R. Yu, E.J. Lee, G. Han and S.-C. Kim, Abstracts Gait analysis of peripheral vestibular loss patients. The 25th ARO midwinter research meeting, 2012, p. 132, poster #895.

Copyright of Journal of Vestibular Research: Equilibrium & Orientation is the property of IOS Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Ambulatory balance monitoring using a wireless attachable three-axis accelerometer.

The ability of conventional diagnostic equipment to monitor feelings of dizziness experienced during daily activities is limited. Our goal is to devel...
611KB Sizes 0 Downloads 0 Views