1 Introduction

1.1 Deep core muscles’ significance

Deep core muscles, as part of lumbopelvic-hip complex (LPHC), play a crucial role in the control of trunk position and motion over the pelvis, helping to optimize force control and motion in the distal segments needed for efficient movement control (Granacher et al. 2013). Core stabilization and spinal stability are achieved through increased intraabdominal pressure and multifidus transversus abdominis, diaphragm and pelvic floor muscle activation (Hodges et al. 2005). Proper breathing influences abdominal pressure which can reduce reliance on accessory muscle contractions. Further, abdominal muscle activation can also increase the diaphragm’s efficiency by ensuring the optimal length and dome shape of the diaphragm (Ko et al. 2018). Weakened or insufficient motor control in the deep trunk musculature can contribute to reduced movement quality both in populations with low back pain and those without low back pain (Emami et al. 2018). Low back pain and altered posture leading to inadequate diaphragm and core muscle activity are characteristic traits for leg amputees (Wasser et al. 2019). This leads to a decline in physical fitness, poor movement quality and an inability to meet the high energy expenditure demands necessary for walking with a prosthesis post amputation (Vestering et al. 2005). Thus, the inactivity and inadequate deep muscle activity eventually may lead to repeated amputations or mortality. Study of the sequence of muscle activation during whole-body movements found that some of the core stabilizers (i. e., transversus abdominis, multifidus, rectus abdominis, and oblique abdominals) were consistently activated before any limb movements (Abiko et al. 2015; Sivapuratharasu et al. 2019).

Pursed lip breathing is a simple technique for slowing down the breathing in order to get more air into the lungs. Pursed lip breathing, one of the deep breathing techniques, is known to increase core muscle activity levels due to abdominal involvement in the process (Tonks et al. 2018; Friel et al. 2005; Ma et al. 2019). Proper breathing technique affects intraabdominal pressure, which reduces the need for compensatory muscle activation.

1.2 Low back pain and lower extremity amputees

The most common types of back pain are non-specific low back pain (NLBP) and chronic low back pain (CLBP). It affects nearly 80% of the world’s population. Most people will experience this type of pain at some point in their lives. While LBP is a common cause of disability in the general population, it is even more common and frequently compounds mobility-related disability in those who have had a lower limb amputated. In a survey of people who had amputations, 52% reported persistent, bothersome low back pain, and 25% reported severe, frequent pain that significantly interfered with their daily lives and activities(Emami et al. 2018).

Due to the limitations of using ultrasound, CT, and MRI technology during movement, there is insufficient information on the effect of deep core muscle exercise in individuals with low back pain in populations with and without lower extremity amputations (Seo et al. 2017; Abdelraouf et al. 2020).

Although there appears to be a link between amputation and low back pain in both populations, the reason for this is unknown(Sivapuratharasu et al. 2019).

Even athletes are prone to chronic low back pain as a result of weak deep core muscles and increased fatigue as a result of superficial muscle compensational mechanisms(Abdelraouf et al. 2020). As a result, obtaining objective data on specific muscle activation is critical.

1.3 Technologies used to increase the efficacy of physical therapy in targeting deep core muscles and low back pain

Because of the complexities of deep core muscle identification and the lack of direct feedback, the effect of the prescribed exercises is unclear. Because of its low cost and ability to monitor activation in multiple locations at the same time, electromyography (sEMG) is an appealing option for feedback during movement. The amplitude of the EMG signal is commonly reported as a raw (in millivolts) or relative value expressed as a percentage of maximum voluntary isometric contraction (%MVIC), and this value provides a general representation of the neural drive to the muscle (i. e., the number of motor unit action potentials generated over a period of time). As a result, physical fitness exercises that elicit more electromyographic (EMG) activity may pose more challenges to the neuromuscular system and, as a result, may be the most effective for improving core muscle strength and stability (Birckhead et al. 2021).

According to European Guidelines for the Management of Chronic Non-Specific Low Back Pain (Debarba et al. 2018; Stamm et al. 2020), no imaging technology, including computer tomography, magnetic resonance imaging, ultrasound, and others, should be used in diagnosing LBP. The mentioned technologies are not discussed in the guidelines as a form of feedback during physical activities to improve treatment efficacy. Electromyography is the only biofeedback technology mentioned in the European Guidelines for the management of chronic non-specific low back pain (EMG). In general physical therapy, as well as in low back pain rehabilitation and prevention programs, targeted activation of deep core muscles and pursed lip breathing training are prescribed.

Real-time feedback necessity. Literature and studies prove the real-time feedback and it’s providing technologies has significant effect on improving exercise and physical therapy efficiency, self-acknowledgement, and higher engagement in activities (Berni and Borgianni 2020; Liu et al. 2020; Blana et al. 2016; Riel et al.2018). The most used physical therapy approaches in amputee rehabilitation, including when treating LBP as a common trait in individuals with LEA, are balance exercises, mobility exercises, proprioception exercises, yoga, proprioceptive neuromuscular facilitation exercises, and stretching (Esposito et al. 2017; Li et al. 2021; Cano Porras et al. 2021; Benady et al. 2021; Applegate et al. 2018).

The technologies used to provide real-time visual feedback are—via mobile applications, computers, large scale monitors based on motion capture, pressure platforms, IMU, ultrasounds, sEMG and other sensor data (Ortegon-Sarmiento et al. 2020; Garcia et al. 2021; García-Bravo et al. 2020a; Kohler et al. 2009). Providing visual feedback related to muscle sEMG and respiration has received little study relative to more costly ultrasound-based measures of muscle thickness (Zhang et al. 2018). Ultrasound and pressure feedback is commonly used as a simple biofeedback system since 1980’ies but it always require assistance from the physician and is limited by the device complexity (Crasto et al. 2019). Ultrasound devices can be used only for in-clinic applications while pressure feedback is suitable for at home use if the patient is appropriately instructed. The pressure biofeedback unit is very limited to a specific kind of exercises and is used strictly as a visual feedback method which does not provide any quantitative data. (Li et al. 2020). Studies which include real-time feedback from muscle activity demonstrate better improvement in LBP exercise efficiency (Yuvarani et al. 2020).

However, due to technical limitations, only a few of the previously mentioned technologies can be used in real-time treatment monitoring while exercising – setups such as ultrasound do not allow freedom of movement, have high signal error, and presence artifacts caused by movement of physiological processes or electric wiring. In most existing technologies, real-time monitoring involves indirect general metrics such as the number of repetitions, intensity level, total time, and others that do not demonstrate the effect of real-time feedback. Even though the existing solutions require looking for a screen or otherwise limit the correct performance of the task – functionality only within a specific range of distances from the base stations (Angelucci and Aliverti 2020), limitations in a 3D space depth, small angles and the capturing area. Movement compensation mechanisms should be taken into account while using this approach.

The majority of solutions have virtual or mixed reality 3D content as general visually guiding feedback without integration with sensor data that alters the 3D environment according to the HMD user, it is common that the sensor data is recorded in a parallel process. The aim of virtual activities for stroke, gait or other rehabilitation (described in the next section) differ from deep core muscle activation physical therapy aims and methods, therefore, requires different technological approach. Accordingly, surface electromyography in combination with the mixed reality headset is proposed in the study as the most suitable combination to ensure unlimited positions for low intensity activities, reduced risk of motion sickness and feedback from the hard to locate deep core muscles thus providing highly individualized solution.

1.4 Virtual reality (VR) and mixed reality (MR)

Even in science and research, the terms virtual, augmented, and mixed reality have been misused. Virtual reality (VR) implies a completely immersive experience that isolates the user from the outside world. Augmented reality (AR) uses the camera on a smartphone to add digital elements to a live view. Real-world and digital objects interact in a Mixed Reality (MR) experience, which combines elements of both AR and VR. The augmented reality solution that corresponds to the mixed reality term is discussed in the following section, whereas augmented reality solutions are not covered in this article. The digital environment ensures the possibility to perform activities that are otherwise not possible or are limited (costs, environment, procedures (surgeries) etc.), to simulate different scenarios, to alter the rehabilitation plan based on objective data – eye movement, locomotion, focus, heat maps and other information possible to obtain from the headset or the applications.

To improve patients' emotional reactions, body image, and physical function, VR and MR have been combined with physical rehabilitation and psychological treatment (Lin et al. 2019). VR and MR could assist manual therapists in overcoming patients’ negative pain beliefs, addressing pain-related emotional issues, and educating them about functional posture and movements. VR and MR could also interact with and change the sensorimotor neural maps that the brain employs to deal with environmental stressors. As a result, combining manual techniques with VR and MR may define a whole mind–body intervention that uses psychological, interoceptive, and exteroceptive stimulations to rebalance sensorimotor integration, distorted perceptions, including visual, and body images. Immersion increases as multi-sensory (e. g., visual, audio, and tactile) experience is combined with simulation via equipment (e. g., a head-mounted display (HMD) or wearable haptic devices) (Tack 2019). The greatest variation and also rehabilitation outcome improvements in rehabilitation solutions has been focused on pain distraction, hand and shoulder rehabilitation (range of motion), posture correction, and neurology, with little or no objective sensor data input (Mfolsé et al. 2014; Gailey et al. 2002; Chen and Angeles 2017; Davidson and Keating 2002). As a result, there is a hazy understanding of the effectiveness of technology. Human compensational mechanisms are a challenging issue; for example, populations with chronic posture alterations frequently have deviated movement trajectories despite the overall “correct” initial position according to headset data(Stamm et al. 2020). VR may be more effective in patients with CLBP who have kinesiophobia (fear of movement due to pain expectation) due to distraction, neuromodulation, and graded exposure therapy (Tack 2019). The preceding discussion focuses on the fear and psychological aspects caused by CLBP rather than the physical therapy used to treat the low back pain.

The content in the mixed reality field has primarily been created for educational and training purposes in anatomy and surgical procedures (Birckhead et al. 2019; Berni and Borgianni 2020; Hersh et al. 2021; Dadario et al. 2021).

In recent years, there has been a significant advancement in the integration of sensors with virtual or augmented/mixed reality solutions specifically for amputees. Most technologies for amputees focus on ADHD, phantom pain rehabilitation (Liu et al. 2020; Bilal et al. 2020; Marinou et al. 2017; Tsai et al. 2018) or prosthetic device control (Blana et al. 2016), while the only physical therapy-related solution is based on the Kinect motion tracking system, which has many technical limitations (Penelle and Debeir 2014; Williams et al. 2019). Amputees have benefited from visual feedback gait training in virtual reality (Esposito et al. 2017).

The reason why there are no mixed reality solutions in physical therapy and exercising is unclear since it might give many benefits such as reduced cognitive load, motion sickness and increased freedom of movement.

In terms of low back pain, there has been a significant increase in progress, particularly when it comes to integrating sensor networks with the immersive environment. Despite the fact that the emphasis is on postural adjustments (Li et al. 2021), patterns of whole-body activation (Cano Porras et al. 2021), and gait training (Benady et al. 2021). In these complex setups, sEMG sensors are increasingly being used in conjunction with accelerometers or force platforms (Applegate et al. 2018). When comparing efficiency in different activities, there are still virtual reality solutions that focus on user experience and pain scale evaluation (Ortegon-Sarmiento et al. 2020; Trujillo et al. 2020; Darnall et al. 2020). Other indirect metrics are also assessed over time, including activity, stress, mood, and sleep (Garcia et al. 2021).

However, there are only a few solutions that focus on physical therapy and exercise (Abdelraouf et al. 2020), such as the researched core endurance of college male athletes. Some progress has also been made toward cardiac rehabilitation (García-Bravo et al. 2020) and remote rehabilitation (Birckhead et al. 2021; Martin Sagayam et al. 2020), both of which are critical for the amputee population.

The aim of the study was to validate a rehabilitation feedback approach for improving recovery after leg amputation, specifically to compare the effectiveness of combined verbal (VF) and visual (MR) mixed reality feedback (intervention cycle) during the exercise cycle to verbal feedback only (control cycle) in patients with lower limb amputation. Muscle electrical activity magnitude and a self-reported questionnaire were used to assess the effectiveness in terms of subjective participant perception of effort, movement control, and muscle electrical activity magnitude. The second objective was to assess the efficacy of a specific exercise program designed to engage deep core muscles by measuring changes in muscle electrical activity magnitude and chest excursion. Anthropometric data, state of health, subjective low back pain (Oswestry Disability Index), and physical activity level (IPAQ) estimation were all examined as factors that could influence exercise and outcomes.

This research is an important step toward developing a system that will allow for individual adjustment (Kolhler et al. 2009) and whole-body monitoring in people with weakness and comorbid conditions (Kaptein et al. 2018; Mundell et al. 2018). From a physical therapy standpoint, the long-term goal is to improve the core strength and functionality of leg amputees; thus, the exercises have been modified to counteract the effects of limb loss, a compromised cardiovascular system, reduced stability, and a compromised vestibular system by providing exercises with low intensity and simple positions (Mundell et al. 2018; Matja Ić and Burger 2003).

The authors hypothesized that incorporating the proposed technological approach to provide real-time feedback during exercise would improve the short and long-term effectiveness of physical therapy. Feedback can help patients become more self-aware and improve the quality of their exercise performance. Obtaining data before, during, and after an exercise session may aid in the development of a patient-centred individual rehabilitation plan, monitoring progress, and increasing the physical therapist’s capability. Understanding the relationship between subjective patient perception and sensor data may help to accelerate and improve technological solution development. The sEMG data will allow for an objective assessment of the effectiveness of targeted deep core muscle exercise (Bilal et al. 2020).

1.4.1 Mixed reality application setup

Data collection and visual feedback during exercises in real-time, the system functionality is developed as follows: the data from sEMG and RIP sensors are transferred to the computer via a Bluetooth connection. The data in the PC are received in the Opensignals software (PLUX – Wireless Biosignals, S.A., Portugal) to detect the signal quality and store the data. Further, only the data from the corresponding channel are transferred to the Hololens 2 mixed reality headset (Microsoft 2021) after filtering and processing in the custom PC application. The operator controls the process, for example – by pausing the data collection, changing the muscle activity range and changing the sequence of the exercises in the headset if necessary. More detailed technical description can be found at (Lancere and Kugaudo 2021).

2 Materials and methods

2.1 Participants

The East Tallinn Central Hospital (Estonia) database was used to recruit 13 patients who had undergone unilateral transtibial or transfemoral amputation. Participants had to meet the following criteria to be considered: men or women aged 30–65 years with or without chronic low back pain and leg amputation surgery performed no more than 3 months prior to the study, with the stump post-operative wound healed Patients were excluded if they had a history of moderate or severe brain injury, prior lower or upper extremity, spinal, or other injuries requiring surgery or a decrease in motor function for more than 6 weeks, which would limit their performance in this study; a disorder or disease of the central nervous system (e. g., stroke, cerebral palsy, spinal cord injury, or other conditions or disease that affects the brain or spinal cord). Other exclusion criteria included high pulmonary and/or cardiovascular risk conditions such as heart vessel stenosis, pulmonary embolism, or others, as well as limited function of the lower and upper extremities due to an injury or neurological/musculoskeletal disorder. Participation in this study is prohibited if you have visual or hearing impairments that limit or could limit your ability to see or hear feedback or comply with instructions given during testing, or if you are pregnant.

All candidates who agreed to participate and met all of the inclusion/exclusion criteria provided written informed consent. The study was approved by the Estonian National Institute for Health Development’s ethics committee (study No. 2280, 10.08.2021.).

2.2 Data collection and procedure

The information was gathered through the use of questionnaires or wireless non-invasive sensors. Data from non-invasive surface electromyography (sEMG) and Respiratory Inductance Plethysmography (RIP) sensors were collected:

  1. (a)

    During the baseline assessment, followed by a pre-exercise evaluation,

  2. (b)

    While exercising in (VF) and (VF + MR) conditions,

  3. (c)

    At the post-evaluation.

The full study protocol (which was previously submitted for publication) included additional pre-post-tests related to breathing capacity and chest mobility, but we are focusing on the visual feedback effect on the deep core muscles in this article.

Anthropometric data, state of health information, subjective low back pain and physical activity level estimation, transversus abdominis and multifidus activation tests were all part of the baseline evaluation. The sEMG, RIP, and a self-reported Subjective Effort Questionnaire (SEQ) were used to assess the efficacy of real-time visual feedback effect (VF + MR) on deep core muscle activation magnitude. Trunk Movement Quality Test, Chest Excursion measurements, the Borg Effort Scale, and the Subjective Effort Questionnaire were used for pre- and post-evaluation.

The intervention was as follows – all participants in the study exercised in two cycles / conditions. The first cycle was performed using verbal feedback (VF), and the second cycle was performed using a mixed reality headset and combined audio and visual feedback (VF + MR). Following the evaluation of baseline parameters, subjects were randomly assigned to one of two study arms in a 1:1 ratio. The study activities in the first branch were (VF + MR) – 15 min rest – (VF), and in the second branch (VF) – 15 min rest – (VF + MR). Given the possibility of limited physical activity and low back pain, the rest period between the two cycles was set at 15 min, despite the fact that in resistance training, 5 min rest is considered adequate (Gentil et al. 2017; Patel et al. 2016; Cancelliero-Gaiad et al. 2014). During exercise and breathing cycles, muscle electrical activity and abdominal expansion were measured to provide real-time input for visual feedback in the MR headset. The exercises have been modified to provide symmetric left and right-side muscle engagement as well as the ability to perform the exercises without the prosthesis. Each cycle begins with practice exercises for Pursed Lip Breathing (PLB) and Transversus Abdominis (TA). The four exercises were randomized between participants and between VF and VF + MR cycles to eliminate the learning and fatigue effect (Wann et al. 2017).

2.3 Methods

2.3.1 Anthropometric characteristics

A standard stadiometer and weight scale were used to measure body height (cm) and weight (kg). The adjusted BMI (kg/m2) was calculated based on the level of amputation (Fadem and Rosenthal 2008).

2.3.2 Chest excursion measurements

Thoracic excursion was determined by subtracting the thoracic circumference at the end of forced inspiration from the thoracic circumference at the end of forced expiration. At the xiphoid process level, use an inelastic measuring tape (Mfolsé et al. 2014; Bockenhauer et al. 2007). The mean value from the 3 measurements was used.

2.3.3 Trunk movement quality test

The sit-to-stand activity was chosen as a common everyday activity to demonstrate core muscle recruitment patterns in order to compare before and after exercise performance (Gailey et al. 2002). This test was used to collect subjective participant feedback on pre-post self-control exercises. The feedback was gathered using the Participant Subjective Effort Questionnaire, in which participants were asked to rate the quality and control of their movements during the sit-to-stand task on a scale of 1 to 10, with 1 being the lowest level and 10 being the highest level.

2.3.4 The Borg Effort scale

The Borg Rating of Perceived Exertion (RPE) enabled measuring the intensity of physical activity during exercises. Following the exercises, the original Borg’s RPE Scale of 6–20 points was used (Chen and Angeles 2017; Penko et al. 2017).

2.3.5 Subjective Effort Questionnaire

The physical load during deep core targeted exercises was evaluated using a The Participant Subjective Effort Questionnaire, and the effect of both sets of exercises was compared. Participants were asked to rank-order the effect of audio vs. visual feedback on breathing or muscle engagement during the session; this questionnaire was designed specifically for this study. The questionnaire included questions about how much the participant engaged his or her abdomen during pursed lip breathing and how much the participant engaged his or her deep abdominal muscles while exercising. In the (VF) and (VF + MR) conditions, the evaluation scale was from 1 to 10 (1 – no visual feedback effect on how much I engage my abdomen in breathing/deep core muscles during exercising; 10 – maximum effect of visual feedback on how much I engage my abdomen in breathing/deep core muscles during exercising).

2.3.5.1 Self-estimated level of possible low back pain

that might influence the subject’s exercise performance and muscle activity levels was assessed by Oswestry Disability Index (Davidson and Keating 2002). Reference score for ODI is 20% and more in people with disability due to low back pain (Tonosu et al. 2012).

2.3.5.2 The International Physical Activity Questionnaire

was used to assess self-estimation of physical activity level (IPAQ; the short version with 7 questions). This provides information about the study population's overall physical activity (Craig et al. 2003).

2.3.5.3 Respiratory Inductance Plethysmography

belt (RespiBAN professional kit (PLUX – Wireless Biosignals, S.A., Portugal (PLUXWirelessBiosignalsS.A., 2021), with RIP respiration belt) and Surface Electromyography (BiosignalPlux, Inc., Portugal) are used to evaluate pre-post effect and to provide real-time feedback. The data was collected using Opensignals software. To detect the presence of deep breathing, the RIP respiration belt was placed between the 10th rib and the umbilicus (Mfolsé et al. 2014). When the patient inhales using his abdomen, the belt stretches and the amplitude of the digital signal increases. The data on amplitude increase were used to provide visual feedback (please see subsection “Combined Audio and Visual Feedback in HMD” for more information).

2.3.6 Surface electromyography

Because the deep core muscles – Transversus Abdominis (TA) and Multifidus (MF) – are the focus of this study, SENIAM guidelines will be followed for sensor placement and signal evaluation (recommendations for sensor locations in trunk or (lower) back muscles). The transversus abdominis muscle is located about 1 cm medial to the anterior superior iliac spine, and the multifidus muscle is located at the level of L5 in the spinous process (about 2–3 cm from the midline). The reference electrode is placed in a bone region, which is electrically neutral, with a range of 1.5 mV and a bandwidth of 25–500 Hz. All of the results were objectively evaluated in order to identify desirable training activities, with a focus on those that resulted in higher activation.

The change in the area under the curve (sEMGAUC) in (VF) and (VF + MR) conditions is used to measure muscle electrical activity. The following data processing will be performed in order to calculate the area under the curve:

  1. (1)

    Butterworth filter (4th order Notch filter, band-pass 50–450 Hz, band stop 45–55 Hz, attenuation 40 dB, passband ripple 3 dB) (Kuo 2021). No band pass filter is used since the sampling rate is 200 Hz. This sampling rate was chosen to provide stable Bluetooth data transfer from 5 channels to the HMD, and literature on post hoc contrasts indicates that there are no significant differences between adjacent sampling rates when using a sampling rate in the 200 Hz range (Larivière et al. 2005; Ives and Wigglesworth 2003). No 50 Hz interference was found in EMG signals (Willigenburg et al. 2010).

  2. (2)

    Root Mean Square filter (window length 100 ms, window overlap 50 ms) (Feldwieser et al. 2012).

  3. (3)

    The length of the muscle activation period is selected.

  4. (4)

    A minimum of three repetitions of the area under the curve value are obtained.

  5. (5)

    The length of the muscle activation during rest is selected.

  6. (6)

    The area under the curve during rest is selected and extracted.

  7. (7)

    The difference between the activation and rest area under the curve is calculated.

  8. (8)

    The average area under the curve in (VF) and (VF + MR) conditions is computed and expressed as a percentage.

2.3.7 Combined audio and visual feedback in HMD

The Hololens 2 headset is used to provide visual feedback. Because there are no controllers and the headset can be used in all exercise positions, the Hololens 2 headset is a wireless device that combines digital objects with the existing environment and allows freedom of movement while exercising. From a physical therapy standpoint, this ensures that exercises are performed in the proper positions.

The audio and visual feedback in the VF + MR cycle is provided only during the exercises. All other tests are carried out without audio or visual feedback. During each exercise, feedback from the sensors is provided only during the active phase, along with a timer for the specific exercise phase, e. g., “lift your hips and hold (5 s, countdown 5-4-3-–2-1)”. When the participant is given instructions, such as “inhale” or “relax”, he or she cannot see any feedback from the sensors.

During the active phase of the exercise, audio and visual feedback is provided in the form of an active circle. Feedback is shown for the respiration/abdominal excursion (the circle fills if the respiration belt data confirms stretching from abdominal breathing) and the muscle electrical activity (the circle fills if the sEMG data confirms deep core muscle contraction during exercising). The sensors provide real-time feedback, allowing participants to recognize the level of their actions and initiate higher levels of engagement.

The user sees the same full pre-exercise instruction and demonstration in a video format in both cycles, (VF + MR) and (VF), but in the (VF) cycle, the commands and count-down are performed by the researcher, and no feedback is received during the exercises. The languages for the (VF + MR) condition are changed based on the patient’s preference – Russian, Estonian, or English.

2.4 Data analysis

In this study, there are two independent variables (IV). The first IV uses data from wearable sensors to provide visual feedback; it has two different cycles – verbal and visual (VF + MR) feedback and verbal (VF) feedback while performing four exercises for deep core muscle strengthening.

The second IV is the type of exercise. The following activities are associated with deep core muscle strengthening: bent leg raise, side plank on both sides, back bridge, and chair lean with resistance band. The primary dependent variable – EMG activation magnitude – was assessed using similar statistical methods within each group, specifically a 2-way repeated measures analysis of variance (ANOVA) to assess the effect of visual feedback on deep core muscle EMG activation magnitude for each muscle across the four exercises. The two-way ANOVA (Feedback X Muscle) was used to compare VF + MR and VF feedback to see how exercise affected the magnitude of multifidus and transversus abdominis muscle activity across the four exercises. Paired t-tests with Bonferroni-holm correction was used for post-hoc comparisons.

Wilcoxon pair-based test with no-parametric distribution was used on all sEMG data to detect the effect of (VR + MR) condition during exercise. The sEMGAUC under the curve (sEMGAUC) was compared for all the population left and right transversus abdominis and multifidus in (VF) condition using Wilcoxon paired test because the exercises were adjusted to activate left and right side symmetrically.

A one-way repeated measures analysis of variance with the Friedman test was used to compare four different exercises in order to identify which ones had the most significant changes.

3 Results

In the Sect. 3.1., the study population characteristics in relation to their physical activity level (IPAQ) and ODI scores are described as well their potential influence on the study results. In the Sect. 3.2., the exercise efficiency, subjective and objective reaction to the exercise program (Figs. 1, 2, 3 and 4) is evaluated. It is important distinguish the efficiency of the exercise program and the visual/audial feedback, to explore the correlation with visual feedback. It was considered that subjective perception might influence the participant’s responses in the self-reported questionnaires concerning the visual feedback, the objective data of sEMG data (Fig. 3) was analysed. For example, high subjective effort because of the character of exercise (Fig. 1) might affect the perception of the technological solution. Also, person’s health state, BMI, physical activity level and potential presence of low back pain can have the effect on exercise performance, not the user experience itself.

Fig. 1
figure 1

The distribution of Borg Effort Scale scores after exercising

Full size image
Fig. 2
figure 2

Trunk Movement Quality test quality and control of movement scores before and after the exercising

Full size image
Fig. 3
figure 3

The comparison of all exercises in terms of sEMGAUC for a Transversus Abdominis (TA) and b Multifidus (MF) muscles

Full size image
Fig. 4
figure 4

sEMGAUC of left and right transversus abdominis (TA) and multifidus (MF) muscles during a Chair lean b Back Bridge c Bent Leg Raise exercises

Full size image

In Sect. 3.3. we demonstrate both subjective (questionnaires, Fig. 5) and objective (muscle electrical activity magnitude, Figs. 6, 7, 8, 9, 10.) data to explore whether there is a correlation between the two. Also, the effect of audial, visual feedback separately on breathing and muscle activation perception is distinguished (Fig. 5a, b, c).

Fig. 5
figure 5

Level of engagement a breathing b muscle activation c The effect of visual and verbally controlled exercise guidance

Full size image
Fig. 6
figure 6

sEMGAUC Chair Lean exercise in (VF) and (VF + MR). TAR – Transversus Abdominis Right side, TAL – Transversus Abdominis Left side, MFR – Multifidus Right side, MFL – Multifidus Left side

Full size image
Fig. 7
figure 7

sEMGAUC Back Bridge exercise in (VF) and (VF + MR) conditions. TAR – Transversus Abdominis Right side, TAL – Transversus Abdominis Left side, MFR – Multifidus Right side, MFL – Multifidus Left side

Full size image
Fig. 8
figure 8

sEMGAUC Bent Leg Raise exercise in (VF) and (VF + MR) conditions. TAR – Transversus Abdominis Right side, TAL – Transversus Abdominis Left side, MFR – Multifidus Right side, MFL – Multifidus Left side

Full size image
Fig. 9
figure 9

sEMGAUC Side Plank Right Side down exercise in (VF) and (VF + MR) conditions for a left leg amputees b right leg amputees

Full size image
Fig. 10
figure 10

sEMGAUC Side Plank Left Side down exercise in (VF) and (VF + MR) conditions for a left leg amputees b right leg amputees

Full size image

3.1 Descriptives

The effect of VR + MR feedback is evaluated in terms chest excursion measurements (indirect measurement of diaphragm activation) – Sect. 3.2, Transversus Abdominis and Multifidus muscle electrical activity magnitude as all being the deep core muscles – Sects. 3.2, 3.3.

The study included 13 participants aged 56 ± 1.9 years (mean value and standard deviation error) with an adjusted BMI of 28.9 ± 1.7 kg/m2 (11 of whom were men, 84.6%), six (46.2%) of whom were transtibial and seven transfemoral amputees. One participant was underweight (BMI < 18,5 kg/m2), five had normal weight (BMI 18,5–24,9 kg/m2), three were overweight (BMI > 25 -29,9 kg/m2) and three obese (BMI > 30 kg/m2). Because of soft tissue elasticity, an increase in BMI may also affect the precision of the chest excursion measurement. Despite the fact that six participants (46.2%) were overweight or obese, sEMG data from the transversus abdominis and multifidus muscles could be recorded.

Although eight of the thirteen participants (61,5%) wore glasses, they could all see mixed reality content. In addition, nine participants (69.2%) reported smoking 5 to 20 cigarettes per day, which may affect cardiopulmonary endurance to perform breathing and physical exercises.

In report of physical activity, three people out of 13 (23,1%) who had lowest IPAQ scores (low, moderate) were obese. Seven participants out of 13 (53,8%) scored a high level of physical activity, this means at least 1.5–2 h of being active throughout the day (Emami et al. 2018). Walking with a prosthetic leg necessitates a greater expenditure of energy. Amputees expend more energy while walking than able-bodied people, so they may overestimate their physical activity level. Even though there was no correlation between the BMI, disability scores and the physical activity level over the population.

In regard to self-reported low back pain, mean ODI scores were 14.3% ± 5.1. Two patients demonstrated 22 and 36% of ODI scores (moderate disability) and two – 42 and 52% score (severe disability).

The physical activity data align with the Borg Rating of Perceived Exertion scores in measuring physical activity intensity level, which showed the average rating of 11.7 (± 0.6) with the majority of answers being 13 (somewhat hard) or bellow (Fig. 1).

3.2 The exercise efficiency

Chest excursion. The mean value of chest excursion before session was 5.2 (± 0.82) cm and after the session – 5.4 (± 0.83) cm. There were noted no significant changes in chest excursion (p = 0.78). For six participants (46%) the chest excursion increased in the range from 0.5 to 3 cm. This could be related to rib cage muscle fatigue, as only two of the participants were familiar with breathing techniques prior to participating in the current study.

Trunk Movement Quality test. After performing the sit to stand activity from the AMPnoPRO battery, 10 out of 13 (77%) the participants confirmed having more control over their body after the exercising (Fig. 2) (p = 0.002). Three people said there was no difference. This data supports the hypothesis that deep core muscles are activated prior to peripheral movement, and that higher deep core muscle activation contributes to better movement quality.

In frames of the study, we compared the exercises for effectiveness in terms of deep core muscle activation magnitude. It is important to determine during which exercises it is possible to engage transversus abdominis (TA) and multifidus (MF) more to achieve faster physical therapy outcome. When targeting to activate Transversus Abdominis muscle (Fig. 3a), the Back Bridge exercise is more effective then Bent Leg Raise (p = 0.046) and Side Plank exercises (p = 0.0013).

To target the multfidus muscle bilaterally (Fig. 3b), the Chair Lean (p < 0.0001) and Back Bridge (p = 0.0041) exercises are more effective than Bent Leg Raise exercise reaching higher area under the curve which means that the participants were able to activate the muscle for a longer period of time. Also, Back Bridge (p = 0.0013) and Bent Leg Raise (p = 0.029) exercises were more effective than Side Plank Right Side and Left side Down respectively. There was no significant difference between other exercises, despite the fact that there was a clear distinction between rest and activation periods during all of the chosen exercises, implying that all of the exercises can be used for deep core muscle activation.

For the Chair Lean, Back Bridge and Bent Leg Raise exercises there was no difference (p > 0.05) between the left and right side sEMGAUC (Fig. 4a, b, c), as a result, we conclude that these exercises are appropriate for engaging deep core muscles symmetrically in unilateral amputees. We also conclude that the data analysis to evaluate the VF + MR feedback effect can be performed considering the mean value of left and right side for the mentioned exercises.

Due to a significant difference between left and right-side TA and MF activation for the Side Plank exercises (Left Side down (pTA > 0.05, pMF = 0.03); or Right Side down (pTA = 0.03, pMF = 0.004)), the visual feedback (VF + MR) effect on muscle engagement for the Side Plank exercise is analysed separately for left (7 participants) and right-side (6 participants) leg amputees. 

3.3 The effect of VF + MR feedback

Self-reported subjective effort questionnaire. The participants indicated that they were making more effort when they saw the visual feedback for pursed lip breathing and muscle activation. In the scale from 1 to 10 (1–there is no visual feedback effect on how much I engage my abdomen in breathing/ deep core muscles during exercising; 10–there is maximal effect of visual feedback on how much I engage my abdomen in breathing/ deep core muscles during exercising), the average scores for the effect on breathing were 6.2 (± 0.3) for engagement without visual feedback (VF) and 7.8 (± 0.5) for engagement using visual feedback (VF + MR). For both there was significant change between (VF) and (VF + MR) conditions – for the engagement of abdominal breathing p = 0.05 (Fig. 5a) and for the effect of muscle engagement p = 0.05 (Fig. 5b). The average scores for the effect on deep core muscle activation were 6.4 (± 0.4) for engagement without visual feedback (VF) and 8 (± 0.4) for engagement using visual feedback (VF + MR). Ten of the thirteen participants agreed that visual feedback has an effect on making an effort to both breathe in the recommended pattern and activate the deep core muscles. Receiving immediate feedback, according to nine out of thirteen participants, has a greater impact on difficulty performing the exercises than performing exercises after a verbal countdown, in other words—controlled exercise guidance (Fig. 5c).

The Muscle Electrical Activity magnitude. For the Chair Lean, Back Bridge and Bent Leg Raise exercises there was no difference between the left and right side sEMGAUC therefore, the (VF) and (VF + MR) conditions were compared for the whole population for these 3 exercises. Even though the change of sEMGAUC after applying the visual feedback is not symmetrical in all exercises (Figs. 6, 7, 8).

Considering all 13 patients included in this study, visual feedback during exercising had a significant effect on TAL activation (p value = 0.002) (Fig. 6). When analysing the percentage of change, the average improvement for (VF + MR) condition in comparison to (VF) condition was 55%. We need to take into consideration that among the participants there were 5 individuals with high improvement, namely, the percentage of change was in the range from 105 to 375% either for TAL or TAR (for example, sEMGAUC – increased from 208 to 597). The effect on MF activation was smaller – there was no effect in average values, but 3 participants showed increase of 70 to 90% (for example, sEMGAUC – increased from 35 to 68).

In frames of the Back Bridge exercise (Fig. 7) there was no significance of visual feedback (VR + MR) effect found even though when analysing the percentage of change, the average improvement for (VF + MR) condition in comparison to (VF) condition was 35% for multifidus muscle. Four participants had the percentage of change was in the range from 53 to 390% (for example, sEMGAUC – from 30 to 147)). The effect on TA activation was smaller–there was no effect in average values, but 3 participants showed increase of 49–107% (for example, sEMGAUC from 75 to 129).

For Bent Leg Raise exercise (Fig. 8) there was no significance of visual feedback (VR + MR) effect found even though when analysing the percentage of change, the average improvement for (VF + MR) condition in comparison to (VF) condition was 19% for multifidus muscle. Four participants had the percentage of change was in the range from 33 to 290% (for example, sEMGAUC – from 14 to 53)). The effect on TRA activation was 29% the percentage of change, but 8 participants showed increase of 30 to 123% (for example, sEMGAUC − from 289 to 643).

Neither for left or right-side leg amputees there was a significant change between exercising with or without visual feedback. When analysing the percentage of change for Side Plank Right Side down exercise (Fig. 9), the average improvement for (VF + MR) condition in comparison to (VF) condition was 42% (TAR) and 9% (TAL) for the left leg amputee group (Fig. 9a), 104% (TAR) and 121% (TAL) for the right leg amputee group (Fig. 9b). For the same exercise for Multifidus muscle there was improvement of 15% (MFR) and 218% (MFL) for left leg amputees; 59% (MFR) and 9% (MFL) improvement for right leg amputees. We need to take into consideration that among the participants there were 4 individuals with high improvement, namely, the percentage of change was in the range from 61 to 340% either for left or right TA muscle (for example, sEMGAUC – from 83 to 353). For the MF activation 2 participants showed increase of 84 to 222% (for example, sEMGAUC – from 113 to 366).

When analysing the percentage of change for Side Plank Left Side down exercise (Fig. 10), the average improvement for (VF + MR) condition in comparison to (VF) condition was 9% (TAR) and 9% (TAL) for the left leg amputee group (Fig. 9a), 9% (TAR) and 17% (TAL) for the right leg amputee group (Fig. 9b). For the same exercise for Multifidus muscle there was decrease of 13% (MFR) and 24% (MFL) for left leg amputees but improvement of 61% (MFR) and 4% (MFL) improvement for right leg amputees. Even though among the participants there were 6 individuals with high improvement, namely, the percentage of change was in the range from 40 to 131% either for left or right TA muscle (for example, sEMGAUC – from 34 to 78). For the MF activation 4 participants showed increase of 41 to 297% (for example, sEMGAUC – from 17 to 70).

4 Conclusions and discussion

The findings of a study that looked into the effectiveness of real-time visual feedback during deep core muscle exercises on muscle activity magnitude before and during exercise in people who had lower extremity amputations have been described. The primary aim of this study was to validate a novel rehabilitation technology for improving recovery after leg amputation, specifically to investigate whether a combined verbal (VF) and visual (MR) mixed reality feedback (intervention cycle) will initiate a greater change in muscle electrical activation magnitude during the exercise cycle than verbal feedback alone (control cycle). We hypothesized that using a mixed reality approach with real-time visual feedback would increase the magnitude of deep core muscle activity compared to exercising without visual feedback.

The second goal was to assess the efficacy of a specific exercise program designed to engage deep core muscles by measuring changes in muscle electrical activity magnitude and chest excursion. Anthropometric data, state of health, subjective low back pain (Oswestry Disability Index), and physical activity level (IPAQ) estimation were all examined as factors that could influence exercise and outcomes.

Descriptives. Despite the fact that the average BMI falls into the overweight category, the results confirm the feasibility of obtaining sEMG data in this population and using it to improve patient self-awareness in terms of locating and activating the deep core muscles.

Based on the self-reported outcome data (Fig. 1) and the lack of a statistical difference in muscle electrical activation magnitudes between left and right muscle engagement, as well as the ability to perform the exercises despite the amputation level, we conclude that the selected deep core activation exercises are appropriate for a wide range of amputees with varying activity levels and body mass indexes. Despite the fact that vital parameters such as heart rate were not recorded, none of the participants dropped out due to dizziness or deterioration of their physical condition.

The Oswestry disability index results revealed a lower incidence of low back pain than expected in people over the age of 40. Even though participants with high ODI scores were able to perform the exercises, this suggests that the exercise program chosen may be appropriate for the leg amputee population. Of course, in order to draw statistically significant conclusions, the number of study participants must be greater, at least 30 according to the calculations (Macedo et al. 2012; Sakpal 2010).

The effect of Mixed Reality real-time feedback. The results of the self-reported questionnaires confirm that the proposed solution significantly increased self-awareness and engagement in the target activities – deep core muscle and abdominal breathing activation. We conclude that the proposed solution has a high potential for everyday use in a physical therapy practice based on a number of findings that demonstrate the importance of motivation in long-term engagement rehabilitation success. A larger sample size is required to substantiate the patient perception statements with sEMG data.

Muscle electrical activity magnitude. We conclude that participants engaged the muscles more in the (VF + MR) condition, but we also believe that due to everyday walking and other asymmetrical movement patterns, the muscles on the amputation side were engaged more to maintain body balance, as during the Chair Lean exercise (Fig. 6). We propose that by implementing bilateral deep core muscles early after amputation, it is possible to gain more symmetrical muscle patterns, which may result in more symmetrical posture, better body balance, and movement quality.

Given that there was no statistical difference between (VF) and (VF + MR) conditions during Chair Lean (Fig. 6), Back Bridge (Fig. 7), Bent Leg Raise (Fig. 8), and for multifidus muscles exercises for Transversus Abdominis and Multifidus muscles, but there were high results among many individuals (sEMGAUC increase in the range of 50 to 400%) and in the self-reported questionnaires, the conclusion is that the study population is small, distribution is too large, and masks the impact of visual real-time feedback.

Even though there was no statistical difference in sEMGAUC between left and right-side muscles, when analysing the effect of visual feedback for left and right-side leg amputees, we found out that for left leg amputees there was a significant improvement of TAL activation in Chair Lean exercise (p = 0.03) and TAL activation in Bent Leg Raise exercise (p = 0.0005) in (VR + MR) condition comparing to (VF) condition. Since these exercises are bilateral and descriptive parameters also didn’t provide any alternative influence, we conclude that this specific group of participants might have had different cognitive/perceptive reaction to the technology.

Furthermore, the attitude and perception of technological solutions is highly individual and is influenced by a variety of factors such as cognitive aspects, prior experience with technologies, and a personal bias. The data from sEMGAUC (Figs. 6, 7, 8) and self-reported questionnaire (Figs. 2 and 5) show a difference between the patient's self-perception and the actual therapeutic effect. As a result, the authors emphasize the importance of reviewing the reliability of patient self-reported questionnaires and re-evaluating the correlation between self-reported questionnaires and objective data, which is now possible due to technological advances. In many cases, this may result in the re-establishment of patient evaluation protocols and rehabilitation planning. The proposed approach would also improve the effectiveness of the physical therapy process during exercise, particularly in difficult cases where difficult to recognize deep core muscles must be engaged.

A complex real-time feedback rehabilitation approach based on mixed reality and wireless sensors can improve recovery after unilateral leg amputation, but the concept can only be applied in practice after additional studies with a larger sample size. When using the preferred application to improve deep core muscle activation, quality of movement, and motivation as soon as possible after amputation surgery, asymmetric muscle activation pattern development may be avoided.

The effects of cognitive function on technology acceptance, as well as changes in breathing characteristics as part of the deep core muscle complex, must be investigated.