Functional assessment of the pelvic floor muscles by electromyography : is there a normalization in data analysis ? A systematic review

1Physical therapist, professor of the course of Physical Therapy of the Instituto Federal de Educação, Ciência e Tecnologia do Rio de Janeiro – Rio de Janeiro (RJ), Brazil. 2Physical therapist, professor of the course of Physiotherapy of the Centro Universitário Barão de Mauá – Ribeirão Preto (SP), Brazil. 3Centro de Reabilitação of the Clinical Hospital of the Faculdade de Medicina de Ribeirão Preto of the Universidade de São Paulo – Ribeirão Preto (SP), Brazil. 4Physical therapist, resident physiotherapist of the Universidade Federal de São Paulo – São Paulo (SP), Brazil. 5PhD of the Department of Gynecology and Obstetrics of the Faculdade de Medicina de Ribeirão Preto of the Universidade de São Paulo – Ribeirão Preto (SP), Brazil. 6PhD of the Department of Internal Medicine of the Faculdade de Medicina de Ribeirão Preto of the Universidade de São Paulo – Ribeirão Preto (SP), Brazil. 88 Functional assessment of the pelvic floor muscles by electromyography: is there a normalization in data analysis? A systematic review Avaliação funcional dos músculos do assoalho pélvico pela eletromiografia: existe a normalização na análise de dados? Uma revisão sistemática Evaluación funcional de los músculos del suelo pélvico por la electromiografía: ¿la normalización existe en el análisis de datos? Una revisión sistemática Aline Moreira Ribeiro1, Elaine Cristine Lemes Mateus-Vasconcelos2,3, Thaís Daniel da Silva4, Luiz Gustavo de Oliveira Brito5, Harley Francisco de Oliveira6


INTRODUCTION
The correlation between pelvic floor muscles (PFM) strength and dysfunctions in this region is still an open field of research, with few published data. Several functional PFM assessment methods are used in clinical practice and research, such as digital palpation, perineometry, surface electromyography (sEMG), dynamometry, and imaging modalities, such as ultrasonography and magnetic resonance imaging 1,2 .
Among these assessment methods, electromyography (EMG) outstands as an alternative method to monitor muscle tone at rest, muscle strength, and muscular endurance, and to obtain data on the normal and abnormal physical functions of the PFM. EMG is indicated as a reliable and objective muscle evaluation method that causes no harm to patients 3,4 .
EMG monitors the electrical activity generated by the depolarization of muscle fibers, as a function of voltage effect over time. It is the algebraic sum of all the signals detected in a certain area and can be considered as an indirect measurement of muscle strength. The electrical activity can be collected either with surface or intramuscular electrodes 1,5 . Because PFMs lie deep to the skin surface and the superficial and deep layers of PFMs have distinct functions, to use surface electrodes adhered to the perineum is not ideal to study the deep layer of the PFMs, as unwanted activity (crosstalk) would inevitably be recorded from the superficial PFMs and potentially from other nearby muscles including the anal sphincter, the gluteus, and the obturator hip musculature. Considering that deep PFMs lie adjacent to the vaginal walls, electrodes positioned against the lateral vaginal walls are a convenient mean to record their EMG activity 6 .
Thus, in clinical practice, surface electrodes (vaginal and anal probes) are more widely used due to the high sensitivity of the perineal region and the skills required in using the needle or fine-wire electrodes 2,7 .
The characteristics of the amplitude and frequency of EMG signal are sensible to intrinsic (type of muscle fiber, depth, diameter and amount of tissue between the muscle and electrode) and extrinsic factors (location, orientation and shape of the area of the electrodes). Therefore, the signal amplitude cannot be analyzed directly 3,5,8 .
The "normalization" is required to analyze and compare EMG signals from different individuals, muscles and acquisition. It is a form of transforming the absolute amplitude values into relative values relating to an amplitude value characterized as 100% 8 . There are several forms of normalization of EMG signal. Among them, the most used is the normalization by maximum voluntary contraction (MVC), in which the reference to the standards is the highest value found among certain muscle contractions in question 9 .
Normalization prevents interference on the intensity of contraction, since they remove the effect of other factors that influence the signal capture. Thus, we can compare different muscles and individuals to the amount of energy produced during a certain contraction only after the process of normalization 5 .
However, no normal standards or references have been established for EMG or differences in the methodology, types of probe and collection protocols used. The way these data are analyzed can make the comparison of studies and systematic assessments of results found difficult.
Thus, the aim of this study was to evaluate the methods of analysis of EMG data in studies that encompass the PFM functional evaluation.

Eligibility criteria
Inclusion and exclusion criteria are: • Inclusion criteria: -Scientific articles published in the last ten years (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014), which aimed to perform pelvic floor muscles (PFM) functional assessment using EMG; -Articles published in English and in full. Design -Full-text articles of cross-sectional studies. Participants -Female adult participants. PFM functional assessment -Performed using sEMG with a vaginal/anal probe; -With a description of how the analysis of EMG data is performed.
• Exclusion criteria: -Studies including children and adolescents; -Studies including men; -Animal experiments; -Treatment with EMG biofeedback, without EMG rating; -Other exams that included EMG (urodynamic, for example); -Modified custom probes; -Review articles, guidelines and other studies with different design.

Information sources and search
To select studies related to the topic investigated, we have performed a computer-aided and manual search between September 2013 and February 2014. The consulted databases included: Medline PubMed, Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews, PEDro, and IBECS. We have limited the search to English-language manuscripts published between 2004 and 2014. Articles published in full and available in the databases were included in this study. Our search strategy was broad, aiming to have as many manuscripts as possible to make a good triage. The following keyword was used: "pelvic floor electromyography". Articles identified in the initial search strategy were selected according to the inclusion criteria.
This review followed the methodological quality parameters of systematic reviews and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria were observed.

Data collection process
A standardized data extraction form was used to collect the following data: authors, year, publication, country of origin, study design, sample, age (years), probe used, data collection protocol and whether the normalization of electromyographic data was performed. This process was performed using two independent raters (AMR and ECLMV ).

Risk of bias assessment
The risk of bias was assessed using the Newcastle-Ottawa Scale (NOS). The original NOS was developed to assess the quality of the observational studies; it contains eight items that analyze three dimensions: selection, comparability and outcome (in cohort studies) or exposure (case-control studies). There is a series of options for each item, which reflects the quality of the studies and is scored by a star: the higher the number of the star, the higher is the study quality 10 .
A study may receive a maximum of one star for each numbered item within the "selection" and "outcome" categories. A maximum of two stars may be given for "comparability". The total score is nine: four stars in "selection", two stars in "comparability" and three stars in "outcome" 10 .

Summary measures
The variables analyzed were: type of probe used in the collection of EMG data; collection protocol; and normalized EMG data. Data was summarized in tables. A qualitative analysis was undertaken due to trial heterogeneity and lack of standardized outcome measures. Therewith, a meta-analysis was not performed.

Study selection
A total of 213 publications were identified from MEDLINE, 221 from PubMed, 51 from CENTRAL and the Cochrane Database of Systematic Reviews, 18 from PEDro, and five from IBECS, resulting in 508 related publications. Of the articles identified, 107 were selected after applying the established inclusion and exclusion criteria. Only 23 articles were included for this review ( Figure 1). Table 1 summarizes the results of the methodological quality assessment.   Luginbuehl et al. 20 Pereira et al. 21 Petricelli et al. 22

General characteristics of the studies
The characteristics of the articles included in this study are listed in Table 2. The sample sizes ranged from nine To investigate whether there are differences in center of pressure displacement, trunk motion, and trunk muscle activity in women with and without SUI during static balance tasks when the bladder is empty and moderately full. (continues)

Characteristics of data collection and analysis of surface electromyography
The type of vaginal probe used varied between the studies. The most commonly used were FemiScan (n=4) and Periform (n=9). One study did not mention the type of vaginal probe used. The EMG data collection protocols also varied. Only one study examined the evaluation of the two separate types of muscle fibers (types I and II). Six studies evaluated the PFM in other positions besides the supine position; and three, during effort (cough) or Valsalva maneuver. Only four studies reported a normalization of the sEMG data for analysis. Band-pass filter of 10 to 500 Hz. Data were sampled by a 12-bit analog-to-digital converter at a rate of 1,000 samples per second, stored on a computer, full-wave rectified and low-pass filtered at 50 Hz to smooth the data.

No
(continues) Vaginal: Periform PFM activity at rest and in 5-s MVC.
The EMG was sampled at a rate of 1 kHz, the cutoff frequency of the low-pass filter (Butterworth, 24dB/ Oct.) was set at 500 Hz. Expecting vibration artifacts in the EMG, no high-pass filter was applied to detect the fundamental frequency of vibration as well as the harmonic content in the EMG signal. This is a signal processer with band-pass filters, cutoff frequencies of 20-500 Hz, instrumentation preamplifier (20-fold gain), differential amplifier with bipolar input, and CMRR>100 dB. Band-pass filter with cutoff frequencies at 20-500 Hz, an amplifier gain of 1,000, and a CMRR>120 dB and a 12-bit analog-to-digital signal converting plate with a 2.0 kHz anti-aliasing filter sampling frequency for each channel was used. Band-pass filter with cutoff frequencies at 20-500 Hz, an amplifier gain of 1,000 and a CMRR>120 dB and an A/D 12-bit analog-to-digital signal converting plate with a 2.0 kHz anti-aliasing filter for each channel and an input range of 5 mV was used. EMG data were bandpass-filtered between 20 and 1000 Hz and sampled at 2 kHz using an AMLAB-based data acquisition system.

No
(continues) Vaginal: Periform EMG activity was recorded prior and subsequent to a postural perturbation in which a 1-kg weight was dropped 30 cm into a bucket held. Subjects were instructed to remain relaxed before loading and to catch the load when it contacted the bucket. In the first ten repetitions the subject held the switch and controlled the timing of the drop of the weight (expected condition). In the subsequent ten trials, subjects wore a blindfold and earplugs and the weight release was controlled by the researcher (unexpected condition).
EMG data were amplified 2,000 times, band-pass filtered between 30  Six body positions: MVC sustained for 5 s and rest for 10 s, with 10-min rest between positions.
The sEMG signal-processing unit consists of one portable analog-to-digital channel of an EMG unit operating at voltage between 0 and 100 V, band-pass filtered at 50 Hz.
No Stüpp et al., 2011 29 Vaginal: Chattanooga Group Three 30-s MVCs. The best of the three contractions was used for the analysis.
Band-pass filter with cutoff frequencies at 20-500 Hz, an amplifier gain of 1,000 and a CMRR>120 dB. An A/D 12-bit analog-to-digital signal converting plate with a 2.0 kHz anti-aliasing filter sampling frequency for each channel was used. The plate had an input range of 5 mV.
No Thompson et al., 2006 30 Vaginal: Periform One contraction for 3 s and a maximal straining Valsalva maneuver. The EMG activity was recorded for 3 s and repeated three times with 1-min rest.
The gain setting was 2,000, and the signal was sampled at 1,000 Hz. The EMG raw data was reduced, rectified, band-pass filtered at 4-400 Hz using a fourth-order zero-lag Butterworth filter (National Instruments).
Yes Thompson et al., 2006 31 Vaginal: Periform One contraction for 3 s and a maximal straining Valsalva maneuver. The EMG activity was recorded for 3 s and repeated three times with 1-min rest.
The amplifier gain was 2,000, and the signal was sampled at 1,000 Hz. The EMG raw data was reduced, rectified, band-pass filtered at 4-400 Hz using a fourth-order zero-lag Butterworth filter (National Instruments).

DISCUSSION
This systematic review aimed to determine how EMG data are analyzed in the functional assessment of PFM. Only seven studies normalized the data, according to the recommendations proposed in the Guide for Use and Interpretation of Kinesiologic Electromyographic Data 33 .
The characteristics of the amplitude and frequency of the electromyography signal are sensitive to intrinsic (muscle fiber type, depth, diameter, and amount of tissue between the muscle and electrode) and extrinsic factors (location, orientation, and shape of the area of the electrodes). Thus, the signal amplitude cannot be analyzed directly 3,5,8 . To analyze and compare electromyography signals from different individuals, muscles, and acquisition modes, it is necessary to "normalize" them, which is a form of transforming the absolute values of the amplitude into relative values related to an amplitude value characterized at 100% 8 .
Normalization methods impede any interference on the intensity of the contraction, as they remove the effect of other factors that influence on signal capture. Thus, it is only after the standardization process that we can compare different muscles and individuals considering the amount of energy produced during a given contraction 5 .
We have several ways to normalize the electromyographic signal. Usually, it is performed by dividing the obtained values by a reference point. The most referenced point in the literature is the normalization by the maximum voluntary contraction (MVC); in this point, a reference is attributed to the highest value found among certain contractions from that muscle. In general, patients are oriented to perform three MVCs and the highest value is recorded. The other contractions of the collected protocol will be percentages of the MVC 9 . Some authors use the mean between two or three MVC as reference value 8 . Another possibility of normalization is to use the maximum peak of the electromyographic signal. Similarly, the value of 100% is attributed to the maximum peak, and all the electromyographic signal is normalized using this value 8 .
As aforementioned, the most common method for the normalization of the amplitude of the EMG signal is to use the MVC. This method quantifies more precisely the relative effort of muscular groups, allowing the comparison between patients with and without neuromuscular dysfunction. It may also be defined as the Muscular Utilization Ratio (%), characterized by the ratio between the mechanic demand imposed during the motor activity and the maximum capacity of the muscular group to perform the activity. This ratio is multiplied by one hundred to obtain a percentage (%) based on the MVC to develop a specific motor task 33 .
According to Soderberg and Knutson 34 , the decision to normalize or not is based on the type of description and whether one of the research objectives is to compare data. If comparisons are made between subjects, days, muscle, or studies, the normalization is required 35 . Meanwhile, if the subjects act independently and the collection is held on the same day, assessing the same muscle without electrode removal, the normalization is not considered necessary. However, we recommend normalization of data, because this step is required in case the results are compared with similar data from other studies in the future.
Besides depending on physiological properties, sEMG is also influenced by non-physiological properties, such as probe configuration (size, shape, how it is applied, and type of filter used for signal detection) 8 . Some studies have assessed the reliability of the comparison between different probes; however, few studies have evaluated this aspect in Brazil.
Ten different probes were used in the studies included in this review, except for those that did not mention the type of vaginal probe used. Data collection protocols also varied, and patients were evaluated in different positions. These are limiting factors for a systematic review of literature that seeks to evaluate the contribution of sEMG in the functional assessment of PFM in a given population. The heterogeneity of the studies hampered their comparison and systematization of data. However, positively, our study opens space for reflection and discussions on the subject to move toward a standardization of the technique used in the PFM functional assessment.
Another important factor worth mentioning is the sample heterogeneity between studies. Functional assessment of PFM of nulliparous, primiparous, and multiparous women; patients with PFM disorders, such as urinary incontinence and pelvic organ prolapse; and women in menopause and postmenopause contribute to the differences in results due to the effects of delivery, mode of delivery, and hormone changes related to aging, for example, on the PFM function. Variations in sample size and study design are also relevant factors that limit the systematization of data due to the distinct methodologies between studies.
The strength of this review is the originality and the analysis of the risk of bias with a specific tool for cohort studies 10 .
As weaknesses of the study we can mention the low methodological quality of the included studies. A considerable amount of the NOS scoring was lost when we analyzed the "comparability" item due to the lack of a control group for comparisons. "Outcome" was another item with a reduced score due to the lack of information on blinding and/or a follow-up long enough so that results may occur. Unfortunately, this was not found for most of the selected results.
The research question within this systematic review is very important. EMG studies are central to evaluate studies on pelvic floor training efficacy, and, as such, comparison across studies with heterogeneity in the methods used to capture sEMG activity is important.
Thus, standardization of guidelines for sEMG data collection, management, analysis, and interpretation of results in the PFM functional assessment is essential. We hope that this review alerts physiotherapists about PFM disorders regarding these issues.