Physical Therapy Korea

Office workers spend more than 6 hours a day sitting down [1] in poor posture, such as forward head posture (FHP), which can occur due to the distance between the trunk and the desk or the height of the monitor [2,3], during long-term sedentary work [4]. FHP is an abnormal posture in which the head is located relatively forward compared to the neck. Thus, the lower cervical spine forms an excessively forward curve, and the upper thoracic spine forms an excessively backward curve [4,5]. Flexed head and neck and FHP, among the poor postures, increase neck muscle activity and cervical spine load, which can cause neck or upper back pain [6-10]. Moreover, 17.7%–63.0% of office workers have experienced neck pain, and 28%–34% have experienced back pain due to these postures over the past decades [6,11]. Besides office workers, students who use computers for a long time also have a high prevalence of neck or back pain [12,13].

Various intervention methods, such as stretching or strengthening exercises, manual therapy, posture correction, electrotherapy, and thermotherapy, are used to reduce work-related musculoskeletal disorders [14]. Posture correction is widely used for ideal sitting postures [15]. Ideal postures include neutral cervical spine postures that help minimize stress, strain, and maximum body efficiency [15,16]. Nevertheless, maintaining an ideal posture under various circumstances may be difficult [17].

The use of passive or active assistive tools can be effective in correcting and supporting posture [15,18]. Several passive assistive instruments, such as office chairs and support pillows, have been developed to maintain an ideal sitting position by postural correction, but they are inconsistent in maintaining postural alignment [18-21]. Besides passive assistive instruments, active assistive instruments, such as biofeedback and electromyography-based feedback devices, that provide feedback and enable active posture adjustment have also been developed; however, they have the disadvantage of being impractical for daily life use because of their voluminous size and unverified long-term effects [15,22,23]. To compensate for these limitations, studies on posture correction using various active wearable sensors, such as inertial measurement unit (IMU) sensors that are comfortable to use, regardless of environmental restrictions, are being conducted [24-26]. Raya et al. [24] have shown that active wearable sensors are reliable for a range of motion measurement. Furthermore, Ailneni et al. [25] and Kuo et al. [26] have shown that wearable sensors are effective in head and neck posture improvement during computer work but may be inconvenient for certain people, such as office workers who wear headphones and work on computers.

To date, IMU sensors, which typically consist of accelerometers and gyroscopes, have been used in several fields, such as navigation and motion tracking [27,28].

Headphones with IMU sensors for office workers can provide posture correction effectively through feedback, detecting postural changes while wearing a headphone without burden. However, to the best of our knowledge, studies on posture correction headphones that provide visual and auditory feedback using IMU sensors have not yet been conducted.

This study aimed to test the intra-rater reliability and concurrent validity of the angular measurement system as a headphone and necklace system for posture correction using IMU sensors. Moreover, to evaluate the intra-rater test-retest reliability of angular measurement of IMU sensors in the headphone and necklace posture correction system (HANPCS) and three-dimensional motion analysis system (3DMAS) angular measurements for neck and upper trunk flexions in the sitting position and to determine the concurrent validity of angular measurement of IMU sensors in a posture correction system for neck and upper trunk flexions in the sitting position compared with a gold standard 3DMAS. We hypothesized that neck and upper trunk flexion measurements using a HANPCS with IMU sensors are as reliable and valid as the 3DMAS.

1. Participants

Twenty-nine healthy adults (18 males, 11 females) aged between 18 and 35 years participated in this study. Table 1 shows the participants’ demographic characteristics in detail. A sample size of 29 was estimated based on previous studies and the Sample Size Calculator [29,30]. The participants had no spinal pain, previous injury, or surgery in the spinal regions and scored < 10% on the Neck Disability Index (NDI) [26,31]. The NDI has been validated and translated into Korean by Song et al. [32]. Participants with any cervical, neurological, or musculoskeletal disorder affecting neck active range of motion or a history of neck or head injury in the last 12 weeks were excluded [31,33]. This study was approved by the Institutional Review Board of Yonsei University Mirae campus (IRB no. 1041849-202204-BM-081-01). All the participants provided written informed consent and completed the NDI survey.

Table 1 . Participant demographics (N = 29).

Variable	Male (n = 18)	Female (n = 11)
Age (y)	23.1 ± 2.8	22.6 ± 3.0
Height (cm)	174.4 ± 4.9	160.5 ± 5.3
Weight (kg)	75.6 ± 4.9	57.9 ± 10.5
Body mass index (kg/m2)	24.9 ± 3.6	22.4 ± 3.5

Values are presented as mean ± standard deviation..

2. Instrumentation

The posture correction system consisted of a headphone, a necklace, and two IMU sensors (CubeMotion, Seedtech Inc.) connected to the computer. The system provides auditory or visual feedback if the angle, presenting the Euler angle, is measured in real-time and exceeds a preset angle threshold. The preset angle threshold could be set up through the system software, and when the angle of inclination of the sensor applied to the headphone and necklace exceeds the preset angle threshold, the computer screen turns off, or a sound comes out of the headphone, so the participant can recognize that the posture of their neck or back has changed.

Euler angles are used to describe the orientation of a rigid body, such as IMU sensors, relative to a fixed coordinate system [34]. Also, Euler angles were measured using IMU sensors, and data were collected using computer software (Figure 1). One sensor was placed on the right side of the headphone to measure neck flexion, and the other sensor that will be applied to necklace was placed on the spinous process of the fourth thoracic spine (T4) where the position of the sensor of necklace to measure upper trunk flexion (Figure 2). Neck and upper trunk flexion measurements were recorded at a sampling frequency of 50 Hz.

Figure 1. Headphone and necklace posture correction system (HANPCS) software program. ‘Fixed Angle’ means the angle threshold value, and if the angle of the sensor exceeds this value, feedback can be given. If the time is set in the ‘Trigger Delay’, feedback can be given after the set time after the angle of sensor exceeds Fixed Angle. The duration of the auditory feedback from the headphones can be set in ‘Sound Times.’ The computer screen is shut down for the set time, if the time is set in ‘Block Screen.’

Figure 2. Process of the movement. (A) Starting position, (B) neck flexion, (C) upper trunk flexion.

The motion capture system (Vicon, Oxford Metrics Inc.) is a highly reliable and valid device for measuring functional motion [35]. Six reflective markers (14 mm diameter) were attached to the headphones’ headband, and the sensor was placed on T4. Kinematic data of neck and upper trunk flexions were recorded at a sampling frequency of 100 Hz using eight motion capture cameras (Vicon, Oxford Metrics Inc.). Kinematic data were processed using MATLAB (MathWorks Inc.).

3. Procedures

All the participants partook in all the sessions. Sessions 1 and 2 were conducted on the same day. Session 3 was conducted at the same time the next day. Each participant partook in three trials in each session, performing neck and upper trunk flexions thrice in each session. Each session ended when three measurements for each movement were completed. The sequence of movements was randomized using Excel (Microsoft Corp.). Before the beginning of each trial, participants sat on a chair with arms on their thighs, with the back as close as possible to the back of the chair, faced the reference point that was made in the wall at the height of their eyes, maintained hip and knee flexions at approximately 90°, fixed the sole of their feet on the floor, and restricted their lumbar movements using an inelastic belt (Mulligan Mobilization Belt, Manualbelt, Balancebody) for the starting position (Figure 3A). The participants practiced neck and upper trunk flexions to familiarize themselves with movements before starting the session. Figure 3B shows the state in which the maximum neck flexion is performed in the starting position, and Figure 3C shows the state in which the maximum upper trunk flexion is performed in the starting position. Calibration of the IMU sensors in the posture correction system was completed and was set to zero degrees by the rater for precise measurement at the starting position, facing the reference point before starting each trial. After the practice and randomization of the sequence of movements, the participants were asked to perform selected movements, such as neck flexion and upper trunk flexion, maintain that state for 5 seconds in the maximum range, return to the starting position, and repeat the selected movement twice. Then, the participants performed another movement thrice in the same manner. The reflective markers and posture correction system remained in the same region. Session 2 was performed after a 10-minute break following session 1. The participants in session 2 performed the same procedure as in session 1. Reflective markers and posture correction system were reapplied by the examiner to participants when they returned 24 hours after session 2 was completed, and session 3 was conducted as in session 1. The 3DMAS and posture correction system were simultaneously used for each trial.

Figure 3. Sensor and marker placement.

4. Statistical Analysis

Participant demographics were analyzed using descriptive statistics. The mean ± standard deviation (SD) of three trials for each movement was calculated for each session. The data from sessions 1 and 2 were used to estimate intra-day reliability. The data from sessions 1 and 3 were used to evaluate inter-day reliability.

To determine the intra- and inter-day test-retest reliability of the IMU sensors in the HANPCS and 3DMAS, the intraclass correlation coefficient (ICC) = [3,k] with a 95% confidence interval (95% CI) was used. k indicates whether the average value measured k times is used [36]. Therefore, ICC (3,3) was used in this study. The standard error of measurement (SEM) and minimal detectable change (MDC) were used to determine the absolute reliability of the measurement. SEM was defined using the following equation: SEM = SD × √(1-ICC), and the lower the SEM calculated, the higher the reliability of the measurement [37]. The MDC with 95% CI (MDC⁹⁵) was calculated as MDC = √2 × 1.96 × SEM to assess the minimal change required to be 95% confident that can be considered to be a change in the participants, not just a change due to an error when a change in the value occurs in measurement [31,38].

To determine the concurrent validity, the agreement of data between the HANPCS and 3DMAS was analyzed using ICC (3,3) with 95% CI and Bland–Altman plots with 95% limits of agreement (95% LOA). The 95% LOA was calculated using the following formula: 95% LOA = mean difference between measurements ± 1.96 SD [39]. All statistical analyses were performed using the IBM SPSS Package for the Social Sciences version 26.0 software (IBM Corp.).

The mean ± SD values of the upper trunk flexion and neck flexion measurements for each session using the HANPCS (Session 1: 30.4 ± 4.9–52.6 ± 8.2; Session 2: 30.7 ± 4.9–52.2 ± 8.0; Session 3: 29.7 ± 4.7–52.2 ± 8.1) and 3DMAS (Session 1: 31.6 ± 4.9–53.6 ± 8.5; Session 2: 32.3 ± 4.9–53.6 ± 8.3; Session 3: 30.5 ± 5.1–52.8 ± 7.9) are presented in Table 2. The intra- and inter-day reliabilities using the two systems for each movement are shown in Table 3 with the ICC, SEM, and MDC⁹⁵ values. All the ICC values are expressed to be above 0.75 for neck and upper trunk flexions using two systems, presenting good to excellent relative reliability. Intra-day (ICC = 0.954–0.971) values of upper trunk flexion and neck flexion using the posture correction system were higher than the inter-day (ICC = 0.865–0.937) values. SEM was low for upper trunk flexion and neck flexion using the HANPCS (Intra-day: 1.05°–1.37°; Inter-day: 1.76°–2.04°) and 3DMAS (Intra-day: 0.84°–1.15°; Inter-day: 1.38°–1.90°). MDC⁹⁵ was also low for upper trunk flexion and neck flexion using the HANPCS (Intra-day: 2.92°–3.79°; Inter-day: 4.87°–5.65°) and 3DMAS (Intra-day: 2.32°–3.18°; Inter-day: 3.82°–5.20°).

Table 2 . Neck flexion and upper trunk flexion for each session using HANPCS and 3DMAS.

Variable	Session 1	Session 2	Session 3
HANPCS
Neck flexion (˚)	52.6 ± 8.2	52.2 ± 8.0	52.2 ± 8.1
Upper trunk flexion (˚)	30.4 ± 4.9	30.7 ± 4.9	29.7 ± 4.7
3DMAS
Neck flexion (˚)	53.6 ± 8.5	53.6 ± 8.3	52.8 ± 7.9
Upper trunk flexion (˚)	31.6 ± 4.9	32.3 ± 4.9	30.5 ± 5.1

Values are presented as mean ± standard deviation. HANPCS, headphone and necklace posture correction system; 3DMAS, three-dimensional motion analysis system..

Table 3 . Reliability of the flexion angle measurement.

Variable	Intra-day reliability				Inter-day reliability
	ICC (95% CI)	SEM	MDC (95% CI)		ICC (95% CI)	SEM	MDC (95% CI)
HANPCS
Neck flexion (˚)	0.971 (0.939–0.987)	1.37	3.79		0.937 (0.865–0.970)	2.04	5.65
Upper trunk flexion (˚)	0.954 (0.902–0.978)	1.05	2.92		0.865 (0.713–0.937)	1.76	4.87
3DMAS
Neck flexion (˚)	0.981 (0.960–0.991)	1.15	3.18		0.946 (0.884–0.974)	1.90	5.20
Upper trunk flexion (˚)	0.970 (0.937–0.986)	0.84	2.32		0.923 (0.835–0.964)	1.38	3.82

HANPCS, headphone and necklace posture correction system; 3DMAS, three-dimensional motion analysis system; ICC, intraclass correlation coefficient; CI, confidence interval; SEM, standard error of measurement; MDC, minimal detectable change..

Table 4 shows the high agreement of upper trunk flexion and neck flexion measurements between the HANPCS and 3DMAS using ICC for each session (Session 1: 0.985–0.993; Session 2: 0.954–0.994; Session 3: 0.960–0.990). The ICC value for all sessions was > 0.90, indicating excellent validity. Scatterplots were presented to express the relationship between the angle measurements of the two systems in Figure 4.

Table 4 . Validity of the flexion angle measurement.

Variable	Session 1		Session 2		Session 3
	ICC (95% CI)		ICC (95% CI)		ICC (95% CI)
Neck flexion (˚)	0.993 (0.984–0.997)		0.994 (0.987–0.997)		0.990 (0.979–0.995)
Upper trunk flexion (˚)	0.985 (0.967–0.993)		0.954 (0.902–0.978)		0.960 (0.915–0.981)

ICC, intraclass correlation coefficient; CI, confidence interval..

Figure 4. Scatterplot using HANPCS and 3DMAS. (A) Neck flexion and (B) upper trunk flexion. HANPCS, headphone and necklace posture correction system; 3DMAS, three-dimensional motion analysis system.

Bland–Altman plots with 95% LOA were used to assess the level of agreement between the two systems and showed the mean difference ± SD values of neck flexion and upper trunk flexion between the two system measurements (Session 1: 0.966° ± 1.405°–1.245° ± 1.136°; Session 2: 1.352° ± 1.279°– 1.541° ± 1.919°; Session 3: 0.534° ± 1.541°–0.841° ± 1.767°) (Figure 5). The 95% LOAs were –1.789° to 3.720° (neck flexion in session 1), –1.156° to 3.859° (neck flexion in session 2), –2.486° to 3.555° (neck flexion in session 3), –0.982° to 3.472° (upper trunk flexion in session 1), –2.219° to 5.302° (upper trunk flexion in session 2), and –2.622° to 4.304° (upper trunk flexion in session 3).

Figure 5. Bland–Altman plot using HANPCS and 3DMAS. (A) Neck flexion in session 1, (B) upper trunk flexion in session 1, (C) neck flexion in session 2, (D) upper trunk flexion in session 2, (E) neck flexion in session 3, and (F) upper trunk flexion in session 3. HANPCS, headphone and necklace posture correction system; 3DMAS, three-dimensional motion analysis system; SD, standard deviation.

This study aimed to examine the intra-rater test-retest reliability and concurrent validity of neck and upper trunk flexion measurements using the HANPCS with IMU sensors compared with the 3DMAS in healthy participants.

This study demonstrates that IMU sensors in the headphone and necklace system for posture correction can be a reliable and valid device as the 3DMAS to measure neck and upper trunk flexions. One of the major findings of our study was that the HANPCS might indicate good to excellent intra- and inter-day reliability in both neck (ICC = 0.937–0.971) and upper trunk (ICC = 0.865–0.954) flexions, with low SEM and MDC⁹⁵ values. In previous studies, to increase the reliability of Cervical Range of Motion measurement using IMU sensors, the rater screened the health status of all participants, excluded those with abnormalities, provided standardized movement sequences and explanations, and performed repeated measurements [24,31]. As a result, previous studies have shown good to high ICC values in all movements ranging from 0.70 to 0.96 [24,31]. However, our findings showed higher ICC values in neck flexion than those reported in previous studies. Potential reasons for the higher reliability were that we minimized the error of measurement by fixing the lumbar region with an inelastic belt, preventing compensatory movements and concurrent measurements to increase the accuracy of measurements. The intra-day ICC value was significantly higher than the inter-day value in the measurement of the two systems for each movement. Previous studies have also reported that intra-day reliability was higher than inter-day reliability [40,41]. Adaptation in the movements caused by repeated tasks can be a possible reason for the higher intra-day ICC values than inter-day ICC values [40,41]. Relatively low intra- and inter-day SEM and MDC⁹⁵ values were demonstrated for the measurement of the posture correction system for each movement, which could indicate proper absolute reliability [42]. Furthermore, the intra-day SEM and MDC⁹⁵ values were lower than the inter-day values in the measurement of the two systems for each movement that could be affected by adaptation from repetitive tasks.

For concurrent validity, we found that the HANPCS had high validity in both neck (ICC = 0.990–0.994) and upper trunk (ICC = 0.954–0.985) flexions in each session, which were excellent values. Simultaneous measurements could be a possible reason for the high ICC values obtained from the results [43]. Small mean differences and less systematic error between the systems are shown in the Bland–Altman plots, and the error from the rater or device could be the potential reason for the differences between the concurrent measurement of the HANPCS and 3DMAS [44]. Also, we could expect the range of difference between the measurement of two systems through 95% LOA in the Bland-Altman plots [45]. If the 95% LOA was within ± 5°, it would be a valid measurement instrument in the clinical circumstances [33].

As a result, our study demonstrates that HANPCS is as accurate in measuring angle as the gold standard, 3DMAS. The system can accurately measure angle changes, provide timely feedback, have fewer setup tasks than the 3DMAS, and is easy to apply in an actual working environment. These findings indicate that HANPCS can be considered a reliable, valid, and accurate device to help office workers correct their poor posture during long-term sedentary work.

This study had some limitations. First, participants were young, healthy, and asymptomatic. Thus, the results of this study may not be relevant to other populations, such as symptomatic or older patients, owing to the inclusion criteria of this study. Second, evaluating inter-rater reliability was necessary because measurements from different examiners could be inconsistent with our results [46]. However, only one examiner participated in this study. Therefore, further studies using a posture correction system could be conducted with symptomatic patients to determine the inter-rater reliability between sessions.

The angular measurement of the HANPCS in the neck and upper trunk flexions was good to excellent and is reliable and excellent, similar to the gold standard, 3DMAS. Therefore, the HANPCS can provide auditory or visual feedback when the angle is measured accurately over time and exceeds a preset angle threshold because of its high angular measurement reliability and validity.

None.

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education (No. 2021R1F1A104792912).

No potential conflicts of interest relevant to this article are reported.

Conceptualization: GH, CY, Seo-hyun Kim, Su-bin Kim, OL. Data curation: GH. Formal analysis: GH, CY, OL. Funding acquisition: CY, Seo-hyun Kim, OL. Investigation: GH. Methodology: GH, CY, Seo-hyun Kim, Su-bin Kim, OL. Project administration: CY, OL. Resources: GH, Su-bin Kim. Software: GH, Seo-hyun Kim. Supervision: CY, Seo-hyun Kim, OL. Validation: GH, CY, Seo-hyun Kim, Su-bin Kim, OL. Visualization: GH. Writing - original draft: GH. Writing - review & editing: GH, Seo-hyun Kim.