Ultrasound Imaging as a Visual Biofeedback Tool in Rehabilitation: An Updated Systematic Review

Rehabilitative ultrasound imaging (RUSI) is used by physical therapists as a feedback tool for measuring changes in muscle morphology during therapeutic interventions such as motor control exercises (MCE). However, a structured overview of its efficacy is lacking. We aimed to systematically review the efficacy of RUSI for improving MCE programs compared with no feedback and other feedback methods. MEDLINE, PubMed, SCOPUS and Web of Science databases were searched for studies evaluating efficacy data of RUSI to improve muscular morphology, quality, and/or function of skeletal muscles and MCE success. Eleven studies analyzing RUSI feedback during MCE were included. Most studies showed acceptable methodological quality. Seven studies assessed abdominal wall muscles, one assessed pelvic floor muscles, one serratus anterior muscle, and two lumbar multifidi. Eight studies involved healthy subjects and three studies clinical populations. Eight studies assessed muscle thickness and pressure differences during MCE, two assessed the number of trials needed to successfully perform MCE, three assessed the retain success, seven assessed the muscle activity with electromyography and one assessed clinical severity outcomes. Visual RUSI feedback seems to be more effective than tactile and/or verbal biofeedback for improving MCE performance and retention success, but no differences with pressure unit biofeedback were found.


Introduction
Motor control exercise (MCE) consists of an exercise-based intervention focused on the activation of deep muscles to improve the control and coordination of these muscles [1]. MCE is widely used since evidence suggests improvements in pain, function, self-perceived recovery and quality of life up to 12 weeks [1]. Several mechanisms, including the lack of stability of the spine, impaired motor control and/or muscle activity patterns, or disturbed proprioception and restricted range of motion, have been proposed for explaining nonspecific spine pain [2]. Motor control exercises aim to restore muscular coordination, control and capacity by training isolated contractions of deep trunk muscles while maintaining a normal breathing and progressing to pre-activate and maintain the contraction during dynamic and functional tasks [3]. Given the difficulty that some patients can perceive during MCE, these exercises are usually performed in supervised sessions providing biofeedback on the activation of trunk muscles for facilitating the awareness and control of these deep muscles' isolated contractions [4].
According to the definition provided by Blumenstein et al. [5], biofeedback refers to external psychological, physical, or augmented proprioceptive feedback that is used

Study Design
This systematic review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [16]. The international OPS Registry registration link is https://doi.org/10.17605/OSF.IO/CNGW4 (accessed on 15 February 2021).

Data Sources
Since a minimum of three databases are needed for adequate systematic reviews [17], we conducted a search in the following electronic literature: MEDLINE, PubMed, SCOPUS and Web of Science databases from their inception to 18 February 2021. Search strategies were conducted with the assistance of an experienced health science librarian and following the guidelines described by Greenhalgh [18]. Search strategies were based on a combination of MeSH terms and key words following the PICO (Population, Intervention, Comparison, Outcome) question: Population: Adults (older than 18 years old) with or without musculoskeletal pain disease.
Intervention: Use of real-time ultrasound imaging as visual biofeedback during MCE to facilitate the MCE performance or retention success.
Comparator: No biofeedback or other biofeedback method. Outcomes: Improvements in muscular function as assessed with imaging methods (US, magnetic resonance imaging or computed tomography) or EMG.
An example of the search strategy (PubMed database) was as follows:

Study Eligibility Criteria
Experimental studies were eligible for inclusion if they (1) evaluated the efficacy of RUSI as visual feedback compared with any other feedback method; (2) used of RUSI for improving muscle function (either as performance or retaining success) of skeletal muscles; (3) included healthy subjects or symptomatic populations, and, (4) were published in English language. Animal studies, observational studies, descriptive studies, review studies, cadaveric studies, published proceedings, and abstracts were excluded.

Study Appraisal and Synthesis Methods
The Mendeley Desktop v.1.19.4 for Mac OS (Glyph & Cog, LLC 2008) program was used to insert the search hits from the databases. First, those duplicated studies were removed. Second, title and abstracts of the articles were screened for potential eligibility by two reviewers. Third, the full text was analyzed to identify potentially eligible studies. Both reviewers were required to achieve a consensus. If the consensus was not reached, a third reviewer participated in the process to reach the agreement for including or not including the study. A standardized data extraction form containing questions on sample population, methodology (intervention, comparator, tasks and muscle assessed), outcomes and results was used, according to the STARLITE guideline [19].
The methodological quality of the included studies was assessed using the PEDro scale [20]. This scale is used to assess the methodological quality of trials and consists of 11 items. The first item (not included in the total score) relates to external validity and the following 10 are used to calculate the final score evaluating the following features: random allocation, concealed allocation, similarity at baseline, subject blinding, therapist blinding, assessor blinding, lost follow-up, intention-to-treat analysis, between-group statistical comparison, and point and variability measures for at least one key outcome. Total PEDro scores between 0 and 3 are considered "poor", 4 and 5 as "fair", 6 and 8 as "good", and 9 and 10 as "excellent" [20].
Finally, a risk of bias analysis for each study was conducted as recommended for systematic reviews [16]. The RoB 2 tool was used to identify the risk of bias in 5 domains: (1) bias due to randomization; (2) bias due to deviations from intended intervention; (3) bias due to missing data; (4) bias due to outcome measurement; and (5) bias due to selection of the reported result [21].

Methodological Quality and Risk of Bias
The methodological quality scores ranged from 4 to 9 (mean: 6.4, SD: 1.4) out of a maximum of 10 points (Table 1). The most consistent flaws were lack of participants (all studies) and therapist blinding (ten studies), concealed allocation (just five studies considered a concealed allocation) and providing point measures and measures of variability (eight studies).
The risk of bias analysis is described in Figure 2. Seven studies showed an overall low risk of bias [22][23][24]27,28,30,31]. However, four studies presented some concerns regarding the measurement of the outcomes and the reported results which should be considered on data interpretation [25,26,29,32].

Transversus Abdominis
Normalized Thickness: Difference between the measurement from each repetition and the basal measure, divided by the basal condition, and expressed in arbitrary units.

Normalized Pressure:
Difference of pressure between each repetition and the basal measure, divided by the basal condition, and in arbitrary units.
Post hoc power = 0.804. Group differences were found (p = 0.006) without interactions (p = 0.994) or repetition effects (p = 0.468). RUSI feedback resulted in larger changes in thickness than the verbal feedback alone (p < 0.05). The bias between thickness and pressure for feedback with and without ultrasonography was 0.0490 and −0.0080 respectively. Significant correlation was not found between pressure measurement and thickness. The lowest minimal detectable changes were achieved by using the ultrasonography feedback. The ability to perform the AHE differed among groups (p < 0.001). During the initial session, 12.5% of subjects in verbal feedback group, 50.0% of subjects in common clinical feedback group, and 87.5% of subjects in RUSI group were able to perform 3 consecutive AHEs. There was a difference among groups in the mean number of trials until performance criterion was reached (p = 0.0006). No differences were noted among feedback groups with regard to the proportions of subjects able to reach the retention criterion.    During contraction, subjects in the experimental group were required to watch the real-time ultrasound imaging and maintain continuous contraction with maximum effort. Images of the right LM at rest and during maximum isometric contraction were acquired. Images of the right TrA muscle were acquired at rest and during the ADIM maneuver.
All participants were firstly given a verbal explanation regarding the purpose and operation procedure of the experiment and the anatomical structure and function of the muscles before the test. Image acquisition for each condition and each time point (Trest, Tc-max, Tc-15 s, Tc-30 s) was repeated three times.

Lumbar Multifidus Transversus Abdominis
Lumbar multifidus thickness Three separate resting ultrasound images were collected immediately after exhalation TrA Thickness ADIM was used to assess the altered muscle thickness associated with a voluntary contraction of the TrA muscle.
No significant differences were found in the thickness of LM at rest (p > 0.999), Tc-max (p > 0.999), and T15 s (p = 0.414) between the two groups. The ability to recruit LM muscle contraction differed between groups at T30 s (p = 0.006), with subjects in the experimental group that received visual ultrasound biofeedback maintaining a relative maximum contraction. No significant differences were found in the TrA muscle thickness at rest (p > 0.999) and Tc-max (p > 0.999) between the two groups. Significant differences of contraction thickness were found at T15 s (p = 0.031) and T30 s (p = 0.010) between the two groups during the ADIM, with greater TrA muscle contraction thickness in the experimental group. Ultrasound imaging Thickness differences between rest and AHE were compared between the two groups.
The difference in internal IO thickness changes between the groups were significant. The differences in EO thickness changes were only significant among the positions. A post hoc analysis of the differences in EO thickness changes among the positions found significant differences between the crook lying and four-point kneeling positions. The TrA thickness changes showed significant interaction between group and position. In the digital examination group, patients performed a full set of supervised exercises guided by digital per anal examination of the external sphincter. In the RUSI group, patients were taught how to contract the anal sphincters while watching the real-time ultrasound display on the monitor screen, and a full set of exercises were performed during each treatment session. In the anal manometry group, Patients were taught how to contract and relax the anal sphincters while attending to the pressures generated in the anal canal, and a full set of exercises were performed during each treatment session.
All participants performed a full set of exercises, consisting of ten five-second sphincter contractions, each at one-second intervals, repeated ten times (a total of 100 contractions). All patients were urged to perform an identical set of exercises twice per day between outpatient visits and were asked to estimate the percentage of exercises they had actually completed.   with 20 s of rest between measurements. After completing the 10 trials in the acquisition phase, all subjects were asked to return in 1 week for follow-up assessments (retention phase).

Ultrasound imaging
To assess multifidus muscle contraction, the difference between the multifidus muscle thickness at rest and during contraction was calculated.
Subjects from both groups improved their voluntary contraction of the multifidus muscle in the acquisition phase (p < 0.001) and the ability to recruit the multifidus muscle differed between groups (p < 0.05), with subjects in the group that received visual ultrasound biofeedback achieving greater improvements. In addition, the group that received visual ultrasound biofeedback retained their improvement in performance from week 1 to week 2 (p > 0.90), whereas the performance of the other group decreased (p < 0.05). Regarding the populations included in the studies, most of them included healthy subjects [22][23][24][25][26][27]29,32] and just three studies included clinical populations, one study included patients with mild-to-moderate fecal incontinence [30], one study included patients with unilateral subacromial pain [28], and one study included patients with chronic low back pain [31]. In general, RUSI visual feedback was a more effective feedback tool than verbal feedback or single manual facilitation for most of the outcomes assessed (e.g., number of repetitions needed to perform correctly the MCE, muscle thickness, or electromyographic activity) considering that procedures were not consistent between studies. However, it seems equally effective as pressure biofeedback units.

Discussion
This systematic review found that RUSI applied as a visual biofeedback tool during MCE seems to be more effective for increasing muscle thickness, muscle activity and target exercise success when compared with verbal or tactile biofeedback. However, the results analyzed from the included studies suggest no additional benefit using RUSI when compared with pressure unit biofeedback. The studies included showed consistent flaws regarding their methodological quality, e.g., participant and therapist blinding, concealed allocation, point measures and measures of variability, which should be addressed in future studies.
To the best of the author's knowledge, the last systematic review assessing the efficacy of RUSI for enhancing the performance and contraction endurance of skeletal muscles during MCE was published in 2007 and, therefore, findings from more recent evidence have not been previously updated [14]. Although our initial aim was to assess how RUSI could improve muscle function, muscular morphology, quality and/or function of skeletal muscles, most of the studies included healthy populations with neither decreased muscle quality nor decreased function. Therefore, although two studies included clinical pain populations, we cannot make definitive conclusions regarding the efficacy of RUSI for improving the mentioned outcomes.
Different comparative biofeedback methods were considered in studies included in this systematic review. Most of the studies included a common clinical biofeedback group (verbal biofeedback and/or tactile feedback) [22,23,[25][26][27]29,31,32] and results seem to be consistent between trials. Comparative analyses showed larger changes in thickness [22,[25][26][27]29,31,32], greater success for exercise performance (greater success ratio and lower number of trials to reach the first successful MCE performance) [23] and greater electromyographic activity [28] for the RUSI biofeedback groups, but no differences for MCE retention at short-term [23]. In the study conducted by Herbert et al. [24], constant (receiving real-time RUSI of successful or unsuccessful muscle activation on the monitor, but without verbal feedback) and variable (receiving delayed feedback after performing the exercise) RUSI feedback were compared. Although both methods sustained the MCE performance success at short-term, the constant feedback group showed superior motor learning at long-term.
Visual RUSI feedback was compared with pressure unit feedback in two studies [25,30]. The results seem to be consistent since Lee et al. [25] found that pressure unit feedback showed no differences for increasing muscle thickness compared with visual RUSI feedback and Solomon et al. [30] found similar improvements in MCE compliance, strength and clinical outcomes. Surprisingly, none of the studies included in this review compared RUSI feedback with other feedback methods (e.g., electromyography or sensitive stimulus). Although this study conducted by Vera et al. [4] was excluded since full-text is not available, their results showed no differences in muscular thickness change with or without sensitive electrical stimulation in addition to the visual RUSI biofeedback.
Although current evidence strongly supports the presence of motor control adaptations in patients with low back pain (LBP), including altered activation timings, lumbopelvic coordination, balance control and kinematics [32], and since MCE is a common form of exercise for LBP management, surprisingly we only identified two studies investi-gating the efficacy of RUSI in clinical populations (unilateral subacromial pain [28] and fecal incontinence [30]), but none included patients with LBP. Healthy population studies are not enough to conclude that visual RUSI biofeedback would obtain similar improvements in LBP populations for facilitating or improving muscular activity since these populations show brain plastic changes of the trunk musculature representation area [33], indicating less fine control [34]. It should be considered that, although MCE is an effective treatment for non-specific LBP, specially indicated for sub-clinical intermediate pain and middle-aged patients [35], low-to-moderate quality evidence showed no additional benefit over spinal manipulative therapy, other forms of exercise or medical treatment in decreasing pain and disability [36][37][38]. Therefore, future clinical trials should include clinical populations for assessing the efficacy of visual RUSI biofeedback for facilitating MCE comprehension, performance and retainment compared with other biofeedback methods.
Finally, there are some limitations of the current systematic review. First, we have only included articles written in English; so, we may have missed some relevant studies published in other languages. Furthermore, we did not include those studies which were unpublished. Secondly, due to the variability of the MCE procedures and in the outcomes, a meta-analysis could not be conducted.

Conclusions
This systematic review found that visual RUSI biofeedback is more effective than common tactile and/or verbal biofeedback for improving MCE performance and retention success in healthy people. There were no clinically important differences between RUSI and pressure unit biofeedback. More high-quality studies with consistent procedures and clinical populations are needed to confirm these findings.