Evaluation of Agreement Between Static
Posturography Methods Employing Tensometers
and Inertial Sensors
R. KOTAS 1, M. JANC2, M. KAMIŃSKI1, P. MARCINIAK1, E. ZAMYSŁOWSKA-SZMYTKE2, AND W. TYLMAN1
1Department of Microelectronics and Computer Science, Lodz University of Technology, 90-924 Łódź, Poland 2Audiology and Phoniatrics Clinic, Nofer Institute of Occupational Medicine, 91-348 Łódź, Poland
Corresponding author: R. Kotas (rkotas@dmcs.p.lodz.pl)
This work was supported by the Project STRATEGMED Innovative System for Evaluation and Rehabilitation of Human Balance from the National Centre for Research and Development under Grant 2/266299/19NCBR/2016.
ABSTRACT This study investigates the agreement between two methods used for the assessment of human balance system. Static posturography based on tensometers, using a commercially available platform, is used as the reference method. An alternative approach is a portable prototype MediPost system that utilises inertial sensors developed by the authors. Both approaches determine the movement of the subject’s centre of mass, quantifying this movement in terms of angular speed. Data for the evaluation of agreement were obtained from 205 subjects, with each subject simultaneously tested with both methods. During the tests, the subject performed a set of standard procedures involving quiet standing in an upright position. In order to verify the agreement between the evaluated methods, the Bland-Altman, concordance correlation and intraclass correlation coefficients were used. In addition, the trajectories of the centre of gravity were compared. The obtained results show good agreement between the verified methods, even though they are based on different physical phenomena.
INDEX TERMS Balance disorders, inertial sensors, static posturography. I. INTRODUCTION
Inability to keep balance of the body may result from deficiencies in many systems of our organism. Data from the National Ambulatory Medical Care Surveys (NAMCS) from 1999 and 2000 and the National Hospital Ambula-tory Medical Care Surveys (NHAMCS) were summed to obtain averaged annual estimates of US healthcare use. Of 979.485 million people reporting to various health cen-ters, 7.443 million patients reported vertigo and dizziness, accounting for 0.8% of visits made during this period [1]. In 2005, Nauhauser published the results of his research, which was based on 4,869 telephone respondents. They show that dizziness occurs in 5% of respondents in the German population [2]. Whilst in Scotland, Hannaford et al., based on their reports on a representation of 15,788 people surveyed, determined the number of dizziness complaints reported at 21% of those respondents [3]. The dizziness and vertigo
The associate editor coordinating the review of this manuscript and approving it for publication was Datong Liu .
TABLE 1.Distribution of falls - 500 falls in 201 people aged 50–90 years [5].
are a part of symptoms and clinical diagnoses of imbalance disorders, which is especially significant problems in an age-ing population. Approximately 30% of the population over 65 have experienced episodes of imbalance that could cause increasing incidents of falling [4]. The fall frequency in the elderly population from Sheldon [5] is presented in Table 1.
Data from epidemiological studies indicated that bal-ance disorders increase the probability of an injurious fall,
meaning that the management of balance system disorders in the population is a critical issue [4]–[6].
Due to the etiological complexity of imbalance and vertigo disorders, highly specialised tests should be carried out to make a diagnosis. Posturography is among the tests used in the battery and is the only one that supplies information regarding the entire balance system.
Force plate posturography is the popular method of balance assessment but is still not normalized. Dynamic posturog-raphy and Sensory Organization Test (SOT, NeuroCom R) protocol is the most accepted. SOT includes 6 conditions: standing on: 1 - stable surface with open eyes; 2 - stable surface with closed eyes; 3 - stable surface with moving surroundings; 4 - unstable surface with open eyes; 5 - unstable surface with eyes closed; 6 - unstable surface with moving surroundings. Although the using of dynamic posturography is the most widespread, its diagnostical utility in vestibular subjects is rather low, about 50% [7]. Moreover, the dynamic posturography is very expensive, not portable and needs a lot of room.
The second protocol used commercially is Modified clini-cal Test of Interaction on Balance (mCTSiB) in which four conditions are performed as following: 1 - stable surface, with open eyes; 2 - stable surface, with closed eyes; 3 – foam pad, with open eyes; 4 – foam pad, with eyes closed. In this test foam pad substitutes the sway-referenced surface. Static posturography mostly is popular in the simplest version, when two tests on stable surface are performed, without using the foam. For vestibular patients the foam test seems to be important from clinical point of view, because when the somatosensory input and vision are cancelled, the vestibular system is responsible for balance maintaining.
Force plate static posturography measures only the static conditions which should be completed by dynamic tests for better functional assessment. Inertial sensor gives the opportunity to implement more variable balance tests [8]. Validation of static posturography based on inertial sensors is the first step to develop a set of clinical tests for static and dynamic balance assessment.
Inertial sensors were used in balance assessment for many years. At the beginning there were some problems with low frequency data acquisition [9]. As shown by Schumann et al. postural sway during quiet standing of sub-jects with vestibular deficits contains a low frequency com-ponent between 0.1 and 0.2 Hz not seen in their healthy counter parts [10]. Mancini et al. revealed, that the most important body sways which differentiate Parkinson disease patients from healthy group were observed in low frequency range below 3,5 Hz [11]. But in 2011 Whitney underlined, that using modern accelerometers with appropriate frequency response solves this problem [9].
There are several papers in which the comparisons between accelerometery and force plate were performed. Whitney et al. [9] estimated the test-retest reliability of the accelerometery during SOT protocol. In healthy subjects they found a significant relationship between the acceleration
measured at the pelvis and the centre of pressure under almost all SOT conditions. Variation of path length between the tests they attribute to the age of participants due to the changing strategies of balance in the older adults. Mancini et al. [11] tested the relationships between the accelerometery and COP measurements in patients with Parkinson Disease only for quiet standing with the eyes open and gaze fixed in straight ahead position. The measurements were simultaneous with force plate recordings. They found the same ability of both tests to differentiate healthy from Parkinson patients. Neville et al. compared the accelerometers and posturogra-phy recordings in small group of ten young healthy volunteers using own protocol, which to some degree is consistent with mCTSiB [12]. Unfortunately, the conditions with eyes open and closed were performed with the feet close together and with hand on hips, which is not in agreement with original posturographic requirements and may influence the results. Freeman et al. tested the original mCTSiB protocol using accelerometers but compared results to SOT conditions [13]. To the best of our knowledge there is no direct compari-son between mCTSiB protocol of static posturography per-formed on force platform and simultaneous accelerometers recordings.
Regardless of the method used for balance assessment, there are some commonly defined terms. Firstly, balance is defined as the preservation of the vertical projection onto the support area (COG) of the body’s centre of mass (COM). The human body is modelled as unity, as an inverted pendulum in which the body is controlled as a single rigid segment that supports a single mass point (COM) [14].
The most common posturographic measure is the cen-tre of pressure (COP), which is the vector sum of the ground-reaction forces applied to the COM. The pressure is measured using a set of tensometers supporting a rigid platform, on which the subject stands.
The COP and COG are identical only when the body expe-riences no acceleration, i.e., is either motionless or moving with constant angular speed. In all other circumstances there is a difference related to the body acceleration: at higher sway frequencies of the body (higher accelerations) neuromuscular response to the COG displacement is larger, resulting in a larger distance between these two points [15]. The inverted pendulum model may be used to compute this difference and therefore to determine COG through measurement of COP [14].
The attractiveness of posturography based on inertial sen-sors lies in its ability to determine COG without the need for the above mentioned, possibly imprecise, computations. In this case the body is still modelled using the inverted pen-dulum concept, however, the tilt angle of the penpen-dulum may be directly obtained from the inertial sensor, which allows straightforward and unambiguous computation of COG. This is one of the reasons why posturography based on inertial sensors was developed and becomes more and more popular. The remainder of this article is organized as follows: Aim of the study is defined in Section II. Section III
describes the physical principles of both methods. Section IV presents the concept and methodology of the conducted research. Section V gives precise information regarding the algorithms used. The results, obtained from both meth-ods based on standard approaches to verify the reliability and agreement between them, is described in Section VI. Sections VII and VIII contain the analysis of the results and discussion, and conclusions, respectively.
II. AIM OF THE STUDY
Inertial sensors have already found use in the monitoring of body positions and allow for the introduction of many diagnostic tests, including static posturography.
The main aims of the presented research were to formulate a quantitative method for assessing human balance system that is competitive with static posturography and to verify the agreement between results obtained simultaneously by both methods. The new method uses a portable MediPost system based on inertial sensors. This MediPost system is a prototype, co-developed by the authors of this study. Standard static posturography, as a recognised method in the study of balance disorders, is considered as a reference method. This study is the first step in the standardisation of new methods of balance assessment based on inertial sensors for healthy subjects and patients with damage of the vestibular system. Such standardisation may lead to a wider adoption of the pro-posed approaches. As they employ portable and cost-efficient hardware, this may widen the scope of balance assessment scenarios, e.g., enabling assessment in telemedicine.
Our study group included subjects in various age and clinical status – to ensure that two measuring methods are comparable for every age, in healthy subjects and patients with peripheral and central vestibular disorders.
III. AVAILABLE MICRO-ELECTRO-MECHANICAL (MEMS)-BASED SOLUTIONS
Inertial sensors are becoming an increasingly more popular balance measuring tool, with most approaches employing a MEMS implementation of sensors [16]–[20]. A solution based on MEMS has already found application in this type of medical research, but it has never been so directly related to the results of static posturography based on tensometers (force platform).
Individual examples also differ in various details that determine the general nature of the solution. For example, as presented in [21], only an accelerometer is used to measure the postural sway. The lack of a gyroscope and magnetometer limits precision and makes it impossible to analyse rotation of the body.
The solution presented in [22] monitors only the head movement. Head movements are triggered by optical stimu-lation (induction of nystagmus). Only a three-axis accelerom-eter is used to monitor head movements.
In comparison to the solution presented in this article, the method described in [23] is a technically similar solution that differs only in the approach to data analysis. The device
is not used for diagnostics but for activity monitoring. The projection of the COM is not determined, with only the acceleration directly analysed. Sensors are placed on different parts of the body (including hands). Only the accelerometer and gyroscope are used (without a magnetometer).
The system described in an earlier patent [24] uses only a three-axis accelerometer. The sensor is placed on the chest (hung around the neck). The patient’s examination is of a long-term nature and the prediction of falls is determined on the basis of the patient’s individual statistics. Data analysis is based mainly on peaks, means and standard deviations of acceleration. This approach is used to recognise the patient’s activities, including walking, sitting and standing.
The solution presented by Sienko et al. [25] is an example of a tool designed to correct the patient’s tilts by using feed-back between the body position (deflection) and the vibrating activator matrix. The device does not have a diagnostic char-acter.
IV. CONCEPT OF THE APPROACH
A. DATA SOURCE
Study group comprised of 205 subjects including two subgroups: 166 vestibular patients (U) and 39 healthy vol-unteers (H). Vestibular patients inclusion criteria were as follows: vertigo and unbalance, signs of central or periph-eral damage of vestibular system in clinical examination and videonystagmography tests. The inclusion criteria of healthy subgroup were as follows: no vestibular or balance symptoms; normal results of physical examination and labo-ratory tests, including caloric test in videonystagmography. The study groups are described in Table 2. All subjects, recruited by the Audiology Clinic in the Nofer Institute of Occupational Medicine (Łódź, Poland), signed an informed consent form to participate in the study, in accordance with the recommendations of the Bioethics Committee.
TABLE 2.Characteristics of participants.
B. PROCESSING PATH 1) TRAJECTORY TEST
In order to compare the compliance of the methods, a prelimi-nary qualitative test was conducted. The subject’s task, while standing, was to move in such a way that his or her COG reproduces a predefined shape. The trial was simultaneously recorded by both systems. Because no calibration procedure was performed, the anteroposterior (AP) plane of the sensor
may be misaligned with the AP plane of the subject. Conse-quently, the resulting shapes may differ in rotation, but this difference should not be treated as an error.
2) MODIFIED CLINICAL TEST OF SENSORY INTERACTION ON BALANCE (mCTSiB)
All subjects underwent a static posturography examination using a NeuroCom RBalance Manager Rsystem and simul-taneously MediPost. The mCTSIB test protocol was used in this study. The quiet standing in an upright position was performed in four standard tests as described in Section I.
Each test (10 s) was repeated in three trials to give a total of 12 trials. Measurements based on inertial sensors (Medi-Post) employed the device attached at the lumbar level (L5),
as shown in Figure 1.
FIGURE 1. Test stand.
C. USED EQUIPMENT
1) POSTUROGRAPH DESCRIPTION AND MEASUREMENT METHODS
NeuroCom RBalance Manager Rsystems are computerised tools for the assessment and rehabilitation of balance and mobility disorders. During the tests, a long force plate (60’’ static dual force plate) and a foam pad (46×46×13 cm) were used. The long force plate performs measurements using four tensometers.
The system provides a standard mCTSIB test, which was used as the reference measurement in the presented research. The mCTSIB protocol consists of four tests (defined in Section B2), which allows the clinician to determine the presence and level of imbalance. The COG deflection is mea-sured as the sway velocity while trying to stand motionless, with a resolution of 0.1◦/s.
2) MEDIPOST DESCRIPTION AND MEASUREMENT METHODS
The focus of the presented research is a MediPost proto-type device (Figure 2). The device relies on a 3-axis iner-tial measurement unit (IMU) to determine its orientation in space. The IMU contains micro-electro-mechanical sys-tem (MEMS) consisting of an accelerometer, a gyroscope and
FIGURE 2. Assembled sensor board and its enclosure 3D project. a magnetometer. This kind of IMU is particularly suitable for measuring low angular speeds (low sway frequencies), as the computation of orientation based on accelerometers does not suffer from accumulated error: at any time, the orientation can be computed as tilt of the IMU relative to the Earth’s gravity vector. This contrasts with the purely gyroscope-based solutions, where the orientation is determined through constant integration of the angular speed – which means that even small measurement errors are accumulated over time and may overshadow actual movements.
In order to eliminate high-frequency noise, the IMU data are low-pass filtered, firstly in the device, and secondly in the PC software.
Further features of the designed solution include:
• Rechargeable battery power supply – working on the battery for at least 2 h;
• Wireless communication with a computer (or mobile phone) in the Wi-Fi standard;
• Small weight and size, similar to a matchbox;
• An enclosure allowing for quick and stable mounting on the subject’s body;
The following key system components were selected:
• IMU sensor: LSM9DS1;
• Wi-Fi radio unit based on the ESP32 system;
• Power supply block based on a TPS63051 voltage con-verter;
• Li-Po battery charger.
Particular attention, due to the conducted research, was devoted to the system’s enclosure, ensuring that it can be fit to the body (subject’s clothes) in a stable way.
The device is synchronised and controlled by a computer programme. The application connects to a predefined Wi-Fi network and then starts to download data from the MediPost via a defined port. The data package is sent 20 times per sec-ond and consists of information taken from all three sensors of the IMU, in three axes (9 values in total).
V. METHODS AND ALGORITHMS
A. ALGORITHM TO DETERMINE ANGULAR POSITION
To determine the angular position of the device, the IMU algorithm of Madgwick was chosen [26]. The operation of
this algorithm is based on the use of quaternions for the representation of orientation in 3D space. The algorithm was slightly modified and is accessible via contact with the authors.
Based on calculated quaternions, the roll, pitch and yaw angles are determined (Figure 3).
FIGURE 3. Rotation axes for the roll, pitch and yaw angles. In the next step, the following values are determined:
• Current angular speed resulting from roll and pitch
rota-tions:
ω = hypot(pitchi− pitchi−1, rolli− rolli−1)
T (1)
where:
hypot(x, y) = q
x2+ y2 (2)
and: pitch and roll are angles, as shown in Figure 3, i is the time-step number.
• Maximum angular speed resulting from roll and pitch rotations;
• Average angular speed resulting from roll and pitch rotations.
Figure 4 presents a simplified scheme showing the principle for determining the coordinates of the projection of the COG.
FIGURE 4. Determining the coordinates of the projection of the COM (a simplified 3D scheme).
VI. RESULTS
The results of the comparison of trajectories are presented in Table 3. The test gives positive results. The shapes of the
TABLE 3.Comparison of trajectories obtained from posturograph with trajectories obtained from medipost. data were collected simultaneously.
obtained trajectories look similar; however, without consider-ing the rotation in the ground plane (it is unnecessary in this type of test).
As noted by Bland and Altman [27], [28], in clinical medicine, the measurements performed on the living organ-ism are constantly changing and their real value is unknown.
This is why continuous improvement and the creation of new and better tools for measurements is required.
To prove the scientific value of the proposed measure-ment method, its results were compared with a classic static posturography method. The results coming from 12 trials (4 tests, each repeated in 3 trials) were aggregated and anal-ysed together. It cannot be expected that the new method will give exactly the same results as the one used so far, but it should be checked how much the results differ. To prove that the new method may be used for diagnosis instead of the current one, the absolute difference between the results should be small enough, so that it does not pose a problem in clinical interpretation. Figure 5 shows the dependence of results obtained with both methods. Figure 6 presents results in the form of a Bland-Altman diagram.
VII. DISCUSSION
It can be concluded that the trajectories obtained by both methods (without taking rotation into account) are consistent for each measurement. In this respect, the initial qualitative effect was achieved.
FIGURE 5. Chart showing dependence of results obtained with posturography (NeuroCom ) and the results obtained with the use of aR
MediPost prototype (all subjects, red line is a trendline).
FIGURE 6. Bland-Altman plot for proposed measurement method in relation to static posturography results by NeuroCom (all subjects, linesR
represent differences expressed in terms of standard deviation (SD): red line: +1.96 SD and yellow line: −1.96 SD).
TABLE 4. Compliance results between MediPost and static posturography by NeuroCom .R
In this article, in order to check the agreement between static posturography based on tensometers and Medi-Post (based on inertial sensors), three analysis methods are used, namely, the Bland-Altman, concordance correla-tion [29], [30] and intraclass correlacorrela-tion coefficients [31]. Values obtained using these methods are presented in Table 4. It should be noted that the literature does not always present a uniform approach regarding threshold values required to pronounce satisfactory agreement. In this respect, interpre-tation of the Bland-Altman coefficient usually follows the original study [27], in which 95% is given as a threshold. In contrast, thresholds for the intraclass correlation coeffi-cient (ICC) are more varied, with some authors stating that values above 0.75 mean excellent correlation [32] while oth-ers push this threshold up to 0.9 [33]. Similar situation occurs with the concordance correlation coefficient (CCC), where Altman [34] suggests 0.8 as a threshold for ‘‘excellent’’, while the commonly cited report by McBride [30] gives a very steep value of 0.99. Regarding the latter, one must remember that the report discussed laboratory analytical tests and their relation to gold standards; in the case of imbalance assessment neither procedures are of analytical precision, nor can the posturography be treated as a gold standard.
Considering the above variations in the interpretation of values, the authors assumed 95%, 0.8 and 0.8 as satisfactory values for the Bland-Altman test, CCC and ICC, respectively. The obtained numbers indicate therefore excellent agreement between the methods.
VIII. CONCLUSION
Both measuring techniques are based on completely differ-ent physical principles. In static posturography the dynamic effects related to ground reaction forces generated by a standing body over a quadrilateral support are measured. In MediPost an inertial measurement unit estimates body kinetics to assess the body position while maintaining balance.
In the entire group the agreement between methods is acceptable. Moreover, both in healthy young volunteers and older vestibular patients the results of measuring are concor-dant, which makes possible using MediPost in subjects with variable age and health status.
The advantages of presented system in relation to the clas-sic static posturographic solution are:
- potentially lower cost of the device (due to much smaller size and absence of any mechanical components); - Possibility of using the system in any place (including
the patient’s home), which allows for continuous moni-toring of the rehabilitation process;
- Possibility of extending the algorithm with other types of tests (exercises);
- Wireless communication and battery power supply. Thanks to the use of IMU, the device may track more complex body movements than classic posturograph, e.g., rotations or movements involving walking. Moreover, as the device can easily be placed on various body parts, additional anal-yses can be proposed, such as tracking head movements, observation of leg position and subject rotation. Because the employed Madgwick algorithm combines data from three types of sensors, the results are more accurate than in solu-tions based solely on accelerometers or gyroscopes.
The presented results show that the MediPost device based on inertial sensors may be used in place of classical posturog-raphy based on force plates. Current work includes adapting the MediPost device to dynamic tests that require calibration of the device in order to determine the initial orientation in space. Further work will include studying the repeatability of the presented method and expanding the system towards the use of more sensors and the evaluation of more diagnostic tests (exercises).
REFERENCES
[1] C. W. Burt and S. M. Schappert, ‘‘Ambulatory care visits to physi-cian offices, hospital outpatient departments, and emergency departments: United States, 1999-2000,’’ Vital Health Statist., vol. 13, no. 157, pp. 1–70, 2004.
[2] H. K. Neuhauser, M. von Brevern, A. Radtke, F. Lezius, M. Feldmann, T. Ziese, and T. Lempert, ‘‘Epidemiology of vestibular vertigo: A neu-rotologic survey of the general population,’’ Neurology, vol. 65, no. 6, pp. 898–904, 2005.
[3] P. C. Hannaford, J. A. Simpson, A. F. Bisset, A. Davis, W. McKerrow, and R. Mills, ‘‘The prevalence of ear, nose and throat problems in the commu-nity: Results from a national cross-sectional postal survey in Scotland,’’ Family Pract., vol. 22, pp. 227–233, Mar. 2005.
[4] L. Z. Rubenstein, ‘‘Falls in older people: Epidemiology, risk factors and strategies for prevention,’’ Age Ageing, vol. 35, no. 2, pp. ii37–ii41, 2006. [5] J. H. Sheldon, ‘‘On the natural history of falls in old age,’’ Brit. Med. J.,
vol. 2, no. 5214, pp. 1685–1690, 1960.
[6] L. Yardley, N. Owen, I. Nazareth, and L. Luxon, ‘‘Prevalence and presen-tation of dizziness in a general practice community sample of working age people,’’ Brit. J. Gen. Pract., vol. 48, no. 429, pp. 1131–1135, 1998. [7] R. P. Di Fabio, ‘‘Meta-analysis of the sensitivity and specificity of
plat-form posturography,’’ Otolaryngol.-Head Neck Surg., vol. 122, no. 2, pp. 150–156, 1996.
[8] S. L. Whitney, G. F. Marchetti, A. Schade, and D. M. Wrisley, ‘‘The sen-sitivity and specificity of the timed ‘Up & Go’ and the dynamic gait index for self-reported falls in persons with vestibular disorders,’’ J. Vestibular Res., vol. 14, no. 5, pp. 397–409, 2004.
[9] S. L. Whitney, J. L. Roche, G. F. Marchetti, C. C. Lin, D. P. Steed, G. R. Furman, M. C. Musolino, and M. S. Redfern, ‘‘A comparison of accelerometry and center of pressure measures during computerized dynamic posturography: A measure of balance,’’ Gait Posture, vol. 33, no. 4, pp. 564–599, 2011.
[10] T. Schumann, M. S. Redfern, J. M. Furman, A. El-Jaroudi, and L. F. Chaparro, ‘‘Time-frequency analysis of postural sway,’’ J. Biomech., vol. 28, no. 5, pp. 603–607, 1995.
[11] M. Mancini, A. Salarian, P. Carlson-Kuhta, C. Zampieri, L. King, L. Chiari, and F. B. Horak, ‘‘ISway: A sensitive, valid and reliable measure of postural control,’’ J. Neuroeng. Rehabil., vol. 9, no. 1, p. 59, 2012.
[12] C. Neville, C. Ludlow, and B. Rieger, ‘‘Measuring postural stability with an inertial sensor: Validity and sensitivity,’’ Med. Devices, vol. 8, pp. 447–455, Nov. 2015.
[13] D. L. Freeman, G. Gera, F. B. Horak, M. T. Blackinton, M. Besch, and L. King, ‘‘Instrumented test of sensory integration for balance: A validation study,’’ J. Geriatric Phys. Therapy, vol. 41, no. 2, pp. 77–84, 2018. [14] D. A. Winter, A. E. Patla, F. Prince, M. Ishac, and K. Gielo-Perczak,
‘‘Stiffness control of balance in quiet standing,’’ J. Neurophysiol., vol. 80, no. 3, pp. 1211–1221, 1998.
[15] D. A. Winter, ‘‘Human balance and posture control during standing and walking,’’ Gait Posture, vol. 3, no. 4, pp. 193–214, 1995.
[16] A. K. Bouke, P. van de Ven, M. Gamble, R. O’Connor, K. Murphy, E. Bogan, E. McQuade, P. Finucane, G. Ólaighini, and J. Nelson, ‘‘Eval-uation of waist-mounted tri-axial accelerometer based fall-detection algo-rithms during scripted and continuous unscripted activities,’’ J. Biomech., vol. 43, no. 15, pp. 3051–3057, 2010.
[17] G. Kamen, C. Patten, C. D. Du, and S. Sison, ‘‘An accelerometry-based system for the assessment of balance and postural sway,’’ Gerentology, vol. 44, no. 1, pp. 40–45, 1998.
[18] M. O’Sullivan, C. Blake, C. Cunningham, G. Boyle, and C. Finucane, ‘‘Correlation of accelerometry with clinical balance tests in older fallers and non-fallers,’’ Age Ageing, vol. 38, no. 3, pp. 308–313, 2009. [19] S. Kim, M. Kim, N. Kim, J. H. Hwang, and G. Han, ‘‘Ambulatory
bal-ance monitoring using a wireless attachable three-axis accelerometer,’’ J. Vestibular Res., vol. 23, nos. 4–5, pp. 217–225, 2013.
[20] M. Alessandrini, A. Micarelli, A. Viziano, I. Pavone, G. Costantini, D. Casali, F. Paolizzo, and G. Saggio, ‘‘Body-worn triaxial accelerom-eter coherence and reliability related to static posturography in unilat-eral vestibular failure,’’ ACTA Otorhinolaryngol. Italica, vol. 37, no. 3, pp. 231–236, 2017.
[21] E. Labat, ‘‘Device for ocular stimulation and detectioin of body reactions,’’ U.S. Patent 2010 228 144 A1, Sep. 9, 2010.
[22] B. Farkašová, ‘‘Daily activity monitoring using a device with accelerom-eters,’’ M.S. thesis, Dept. Cybern., Fac. Elect. Eng., Czech Tech. Univ. Prague, Prague, Czechia, 2015.
[23] B. Najafi, A. Vaziri, and A. Boloori, ‘‘Fall detection and fall risk detection systems and methods,’’ U.S. Patent 2016 100 776 A1, Apr. 14, 2016. [24] K. H. Sienko, M. D. Balkwill, and C. Wall, ‘‘Biofeedback improves
postu-ral control recovery from multi-axis discrete perturbations,’’ J. Neuroeng. Rehabil., vol. 9, no. 1, p. 53, 2012.
[25] S. O. H. Madgwick, A. J. L. Harrison, and R. Vaidyanathan, ‘‘Estimation of IMU and MARG orientation using a gradient descent algorithm,’’ in Proc. IEEE Int. Conf. Rehabil. Robot., Zurich, Switzerland, Jun. 2011, pp. 1–7. [26] J. M. Bland and D. G. Altman, ‘‘Statistical methods for assessing agree-ment between two methods of clinical measureagree-ment,’’ Lancet, vol. 327, no. 8476, pp. 307–310, 1986.
[27] J. M. Bland and D. G. Altman, ‘‘Measuring agreement in method compari-son studies,’’ Stat. Methods Med. Res., vol. 8, no. 2, pp. 135–160, Jun. 1999. [28] I. Lawrence and K. Lin, ‘‘A concordance correlation coefficient to evaluate
reproducibility,’’ Biometrics, vol. 45, no. 1, pp. 255–268, Mar. 1989. [29] G. B. McBride, ‘‘A proposal for strength-of-agreement criteria for Lin’s
concordance correlation coefficient,’’ Nat. Inst. Water Atmos. Res., Auck-land, New ZeaAuck-land, NIWA Client Rep. HAM 2005-062, 2005.
[30] P. E. Shrout and J. L. Fleiss, ‘‘Intraclass correlations: Uses in assessing rater reliability,’’ Psychol. Bull., vol. 86, no. 2, pp. 420–428, 1979. [31] D. V. Cicchetti, ‘‘Guidelines, criteria, and rules of thumb for evaluating
normed and standardized assessment instruments in psychology,’’ Psychol. Assessment, vol. 6, no. 4, pp. 284–290, 1994.
[32] T. K. Koo and M. Y. Li, ‘‘A guideline of selecting and reporting intraclass correlation coefficients for reliability research,’’ J. Chiropractic Med., vol. 15, no. 2, pp. 155–163, 2016.
[33] D. G. Altman, Practical Statistics for Medical Research. London, U.K.: Chapman & Hall, 1991, pp. 285–288.
