AN ELASTIC EXERCISE BAND MOUNTED WITH A BANDCIZER™ CAN DIFFERENTIATE BETWEEN COMMONLY PRESCRIBED HOME EXERCISES FOR THE SHOULDER

Abstract

Background

Home‐exercise is commonly prescribed for rehabilitation of the shoulder following injury. There is a lack of technology available to monitor if the patient performs the exercises as prescribed.

Purpose

The purpose of this study was to investigate the validity of using three dimensional (3D) gyroscope data recorded with the Bandcizer™ sensor to differentiate between three elastic band exercises performed in the shoulder joint: abduction, flexion, and external rotation.

Design

Concurrent validity study.

Methods

This study was performed over two phases. In the first phase, 20 subjects performed three sets of 10 of shoulder abduction, external rotation and flexion exercises with a Thera‐Band mounted with a Bandcizer, while supervised by a physical therapist. The Bandcizer has an inbuilt three‐dimensional gyroscope, capable of measuring angular rotation. Gyroscope data were analyzed in Matlab, and a one‐way ANOVA was used to test for significant differences between each of the three exercises. An algorithm was then created in Matlab based on the exercise‐data from the gyroscope, to enable differentiation between the three shoulder exercises. Twenty new subjects were then recruited to cross‐validate the algorithm and investigate if the algorithm could differentiate between the three different shoulder exercises.

Results

A blinded assessor using the Matlab algorithm could correctly identify 56 out of 60 exercise sets. The kappa agreement for the three exercises ranged between 0.86‐0.91.

Conclusion

The ability to differentiate between the home exercises performed by patients after shoulder injury has great implications for future clinical practice and research. When home exercises are the treatments‐of‐choice, clinicians will be able to quantify if the patient performed the exercise as intended. Further research should be aimed at investigating the feasibility of using the Bandcizer™ in a home‐based environment.

Word count

2429

Level of Evidence

2

INTRODUCTION

Home‐based exercises are often a vital part of rehabilitation for patients with shoulder impingement¹ leading to improvements in pain intensity, quality of life and muscle strength, similar to supervised exercises.^2,3 Home‐based exercises are often performed with inexpensive, adjustable and transportable elastic exercise bands, which require little space compared to exercise machines.⁴ Despite the effectiveness of home‐based exercises in shoulder rehabilitation, patient adherence to exercise programs has proven to be a challenge. Life‐long adherence to exercise is required for many patient groups, however in many cases patient adherence is inadequate.^5,6

It is often unknown if a patient's lack of treatment response to an exercise program is due to reduced adherence, or if it is due to a problem with the exercise dose, type of exercise or simply because the patient does not perform the exercises correctly.^6,7 Reliable and valid methods for quantifying adherence to home‐exercise programs are sparse. One method widely used is self‐reporting.⁷ This method is often unreliable, as patients can overestimate or underestimate how much exercise they have done.⁶ Even if the patient has been adherent to an exercise program, it is often uncertain whether the patient has performed a shoulder exercise as they were instructed.⁸ For example it may be relevant to discern whether a patient performed abduction or a different shoulder movement (e.g. shoulder extension or flexion), which would the likelihood that there will be optimal loading of the intended lateral muscle‐tendinous structures of the shoulder (for example: the supraspinatus).

A new commercially available sensor has been developed, which automatically records and stores exercise data on a memory card or onto a mobile phone (the Bandcizer ™). The University of Southern Denmark and the National Danish partnership UNIK developed Bandcizer™. This new sensor may enable clinicians and researchers to measure adherence to home‐based shoulder exercises.^9-11 Bandcizer™ is capable of automatically stamping training data with time‐of‐day and date, quantifying important aspects of exercise quality such as range of motion, and quantifying important exercise descriptors such as time under tension (TUT), based on the Bandcizer's™ recordings of changes in elastic band thickness.⁹ What it lacks, however, is the possibility of determining which home‐based exercises the patient has in fact performed while unsupervised. As the Bandcizer™ also contains a gyroscope, it is likely that gyroscope‐data can be used to differentiate between commonly performed shoulder exercises, thus giving important information on which shoulder exercises/actions the patient actually performed at home.

The purpose of this study was therefore to investigate the validity of using three dimensional (3D) gyroscope data recorded with the Bandcizer™ sensor to differentiate between three elastic band exercises performed in the shoulder joint: abduction, flexion, and external rotation. The authors’ hypothesized that it would be possible to differentiate between each of the three shoulder exercises using data from the 3D gyroscope.

MATERIAL AND METHODS

Ethics

All participants included in the study received verbal and written information regarding the study. Written informed consent was obtained in accordance with the Declaration of Helsinki. The local Ethics Committee approved the study (2012‐2410).

Design

The design was a concurrent validity study, which investigated using the inbuilt 3D gyroscope in the Bandcizer™, attached to a standard elastic exercise band, to differentiate between three commonly performed shoulder exercises: abduction, flexion and external rotation. A physical therapist supervised the proper execution of the three shoulder exercises, ensuring that, for example, shoulder abduction was in fact performed as shoulder abduction. The study was designed with two Phases. Phase one involved collecting data from 20 participants, who each performed three sets of 10 exercise repetitions for each of the three exercises, with the Bandcizer™ attached to a Thera‐Band. The position of the Bandcizer™ was standardized at 5cm from the Thera‐Band handle.

Gyroscope data from the first Phase was used to create a Matlab algorithm with parameters designed to differentiate between the exercises. Phase two included collecting data from 20 new participants who each performed one set of 10 repetitions of shoulder abduction, flexion and external rotation, giving a total of 60 data sets. A blinded assessor was used then used to cross‐validate the algorithm. The reporting of the study follows the Guidelines for Reporting Reliability and Agreement Studies (GRRAS).¹²

Participants

A convenience sample of 20 healthy volunteers (13 females), aged 21‐48 without injury or operation to the shoulder, neck or back participated in the first phase to develop the algorithm and another 20 volunteers aged 20‐34 (13 females) also without injury or operation in the shoulder participated in the second phase of the study to cross‐validate the developed algorithm. Participants were recruited from the student population of two local universities. The sample size in Phase 1 was based on getting at least 60 sets for each of the three exercises to allow Matlab to calculate an average dataset for each exercise based on data from the gyroscope. The sample size in Phase 2 was based on the amount of training a patient would perform during two weeks of home‐based training, having received an initial supervised session with a physical therapist/rehabilitation specialist.

The Bandcizer™

The Bandcizer™ measures the changes in thickness of the elastic band and has previously been validated for measuring TUT.⁹ The Bandcizer™ is equipped with a LSM330 3D digital gyroscope with a sample rate of 20 hz. The data recorded by the sensor is sent directly to a computer via Bluetooth in the form of text files.

Bandcizer's™ 3D gyroscope measures rotation and angular velocity occurring around three axes: x, y and z. The positions of these axes are constant in the Bandcizer™ and cannot be altered. If rotational movement occurs, for example around a vertical axis, then the x‐axis of the gyroscope will register and record this movement, under the condition that the x‐axis of the gyroscope is always positioned on the Thera‐Band to point vertically. This was an independent study, where none of the authors are affiliated with Bandcizer™.

Test setup

A green Thera‐Band was used in this study. The placement of Bandcizer™ on the Thera‐Band was standardized at 5cm from the handle and the 3D gyroscope was always in the same position. This was done by positioning the x‐axis in a vertical direction.

The exercises

The physical therapist responsible for observing the exercises determined the participants 12 repetition maximum (12RM) for each exercise, before data collection took place. The tension on the Thera‐Band was adjusted according to each participant's 12RM. The participants were then asked to perform three sets of 10 repetitions resembling a normal training load in the clinic.

Shoulder abduction, flexion and external rotation were performed with a three second concentric, two seconds isometric and three second eccentric contraction phase. There was a two second break between each repetition where there was no tension in the exercise band.

To assist the participants in maintaining the correct time under tension during the exercises, the participants were shown a video of each exercise before testing took place. The same video also played during the exercises to assist with correct time under tension. A mirror was also provided to assist the participant in reaching the required range of motion, as shown in Figure 2.

Figure 2.

Exercises used in the research, A=Abduction start position, B = Abduction end position; C=Flexion start position, D=Flexion end position; E=External rotation start position, F = External rotation end position.

Figure 1.

Exercise setup with Bandcizer™ placement on Thera‐Band.

A two‐minute break was held after each set of ten repetitions. If the exercises were not performed correctly, the physical therapist would stop the test, allow the participant to take a break, and instruct the participant in the correct technique. Data from exercise sets not completed correctly were deleted, meaning that only correctly performed exercise data was collected and stored.

Data analysis

Data from the three gyroscope axes: x, y and z, was imported to Matlab (MATLAB and Statistics Toolbox Release 2013b, The MathWorks, Inc., Natick, Massachusetts, United States) in the form of text files. Each gyroscope axis was plotted separately as a function of time in Matlab, for every exercise set; a total of 180 plots. Data was filtered using a zero‐lag, 4^th order Butterworth IIR digital low‐pass filter with a frequency of 5Hz. Plots were generated from the data in its filtered form and as a cumulative sum for each axis of each exercise set. The plots were assessed visually to screen for the parameters most likely to differentiate between the exercises. The minimum and maximum angular rates for the raw data and cumulative sum of the data for each axis appeared different between the exercises. These differences were then tested in SPSS using one‐way analysis of variance, and Tukey's test. There were two parameters, which were able to differentiate between the different exercises. These parameters were the minimum angular rate from the x‐axis, which was significantly different between abduction and rotation (p < 0.001) and flexion and rotation (p < 0.001) and the cumulative sum maximum angular rate from the z‐axis, which showed a significant difference between abduction and flexion (p < 0.001). These parameters were used to create a Matlab algorithm using conditional statements that could differentiate data from each of the exercises. The Matlab algorithm was then tested on data from 20 new participants (Phase 2).

The Matlab algorithm is attached as an appendix and is free to use. (Appendix 1)

Statistical analysis

Cohen's unweighted kappa statistic was used to calculate the agreement between the exercises performed under supervision by physical therapist and the answers that the Matlab algorithm gave for the new data.

RESULTS

The kappa agreement between the exercises performed under supervision by physical therapist and the answers that the Matlab algorithm gave was 0.90 (95%CI: 0.81‐0.99) (Table 1). The algorithm correctly identified all 20 external rotation exercises but interpreted one flexion and one abduction movement each as external rotation. Eighteen out of 20 abduction and flexion exercises were correctly identified.

Table 1.

Exercises performed in the second phase, compared to the answers returned by the Matlab algorithm from Phase 1, when used to differentiate between the exercises. The table depicts the number of times each exercise was identified by the algorithm.

Answers Provided by Matlab Algorithm
Exercise	Rotation	Abduction	Flexion
Rotation	20	0	0
Abduction	1	18	1
Flexion	1	1	18

DISCUSSION

Patient adherence to home exercise programs is a vital step towards successful home‐based rehabilitation after shoulder injury, yet adherence in many patient groups is inadequate.^5,6 The Bandcizer™ has been proven to be a valid method for measuring time under tension,⁹ yet it was unknown if the inbuilt 3D gyroscope could be used to determine which type of movement direction the patient actually performed during their home‐based unsupervised exercises. This study investigated the validity of using 3D gyroscope data recorded with the Bandcizer™ sensor to differentiate between three elastic band exercises performed in the shoulder joint; abduction, flexion and external rotation. The results showed that the algorithm could correctly differentiate between 56 of the 60 exercises, corresponding to a kappa agreement of 0.86‐0.91, demonstrating almost perfect agreement.

Practical usage of Bandcizer™

By standardizing the position of the Bandcizer™ sensor on the Thera‐Band, it was possible to distinguish which type of shoulder movement was being performed, due to the corresponding movements on the relevant axis. Using the Bandcizer™ it is possible to measure time under tension, range of motion and differentiate between three commonly used shoulder exercises.^9-11 In a practical scenario, the physical therapist would give the patient an initial instruction in the relevant exercises. This could for example be three sets of 12 repetitions for shoulder abduction and external rotation with a time under tension consisting of three seconds concentric, two seconds isometric and three seconds eccentric phases.¹³ The patient would then perform these exercises at home for two weeks, with the Bandcizer™, and then come to a follow up appointment with the physical therapist. During the follow up appointment, the therapist could review the exercise data and evaluate the patient's adherence to the training program and check if the shoulder abduction was indeed performed as abduction and not as shoulder flexion instead. Data from the Bandcizer™ can be combined with the patient's own reports of changes in pain and function, providing the physical therapist with a strong rationale for how the training dosage should be changed during the next phase of home‐based training.

The algorithm developed for this study was created in Matlab, a software program that due to its expense, may not be a realistic tool for physical therapists to acquire. The principles used to develop the algorithm can however be easily adapted to other programming languages for example freeware such as Python. It would take the physical therapist approximately five minutes to apply the algorithm to the relevant data sets and read the results of the exercises performed. In order to combine this with training dosage (time under tension, number of repetitions and number of training sets it would take an additional 90 seconds per exercise set (see reference⁹ for a more detail).

Limitations

The three shoulder exercises tested in this study are performed across three different axes, allowing for successful differentiation when the positioning of the Bandcizer™ is standardized. This is likely to make it easier to differentiate between the three exercises. Challenges will arise when trying to differentiate exercises, performed on the same axis, for example flexion and extension of the shoulder. In reality a patient is unlikely to perform a shoulder abduction instead of a flexion but rather a combination of the two exercises. This situation would call for a more advanced approach to differentiating between the exercises, where the gyroscope data would need to be more precise than that which can be recorded when attached to an elastic band.

3D gyroscopes have previously been used to assess ROM, gait and functional movement in several previous studies ^14-17 and are usually attached directly to limbs or other body parts. Attaching the gyroscope directly to the body provides more precise data regarding movement, detecting even small variations ¹⁷ compared to that which can be detected when the gyroscope is placed on an elastic band. Despite the potential for tracking movement using 3D gyroscopes, it is not yet possible to combine the quantification of time under tension with more precise measurements of movement using the Bandcizer™, due the fact that Bandcizer™ must be placed on an elastic band.

Further research

The current study was performed under controlled supervised conditions. It can be expected that without supervision, a patient would perform the exercises with a larger variation in the quality, making it difficult to differentiate between the exercises. This could be, for example, because of the placement of the Bandcizer™ on the elastic exercise band. Future research should investigate if it is possible to differentiate between different exercises when the patient performs the exercises at home without supervision.

CONCLUSION

By standardizing the positioning of Bandcizer's™ 3D inbuilt gyroscope, it was possible to create an algorithm that could correctly distinguish between shoulder abduction, flexion and external rotation exercises, by correctly identifying 56 out of 60 exercise sets. The ability to differentiate between the home‐exercises performed by patients after shoulder surgery has great implications for future clinical practice and research, where home exercises are the treatments of choice, as they will enable clinicians to quantify whether the patient performed the exercise as intended.

APPENDIX 1