Wearable Monitoring Devices for Biomechanical Risk Assessment at Work: Current Status and Future Challenges—A Systematic Review

Background: In order to reduce the risk of work-related musculoskeletal disorders (WMSDs) several methods have been developed, accepted by the international literature and used in the workplace. The purpose of this systematic review was to describe recent implementations of wearable sensors for quantitative instrumental-based biomechanical risk assessments in prevention of WMSDs. Methods: Articles written until 7 May 2018 were selected from PubMed, Scopus, Google Scholar and Web of Science using specific keywords. Results: Instrumental approaches based on inertial measurement units and sEMG sensors have been used for direct evaluations to classify lifting tasks into low and high risk categories. Wearable sensors have also been used for direct instrumental evaluations in handling of low loads at high frequency activities by using the local myoelectric manifestation of muscle fatigue estimation. In the field of the rating of standard methods, on-body wireless sensors network-based approaches for real-time ergonomic assessment in industrial manufacturing have been proposed. Conclusions: Few studies foresee the use of wearable technologies for biomechanical risk assessment although the requirement to obtain increasingly quantitative evaluations, the recent miniaturization process and the need to follow a constantly evolving manual handling scenario is prompting their use.


Background
In recent years, wearable sensors have been used for quantitative instrumental-based biomechanical risk assessments in the prevention of work-related musculoskeletal disorders (WMSDs). Previously, in the attempt to reduce the risk of WMSDs while handling materials, handling people in the healthcare sector or while maintaining fixed postures, several methods have been developed, accepted by the international literature and used in the workplace. These approaches have without doubt facilitated prevention activities during the last decades by improving occupational health and safety of people at work but, on the other hand, need a significant update based on two main aspects. First, the standardized methods commonly used for biomechanical risk assessment are still mainly based on observational and subjective approaches [1][2][3][4] and don't include instrumentation-based tools. Second, the recent widespread use of robots, automation and mechanization in industry for the reduction of the physical effort has modified manual handling work activities. One of the key technologies

Results
From the database search, after removing duplicates, 7221 references were retrieved and screened for eligibility based on their titles (Figure 1). Following that, 357 abstracts and 106 full text articles were assessed for eligibility. Articles were excluded due to the facts that wearable devices were not used in biomechanical risk assessment or they were not used in work activities,

Results
From the database search, after removing duplicates, 7221 references were retrieved and screened for eligibility based on their titles ( Figure 1). Following that, 357 abstracts and 106 full text articles were assessed for eligibility. Articles were excluded due to the facts that wearable devices were not used in biomechanical risk assessment or they were not used in work activities, biomechanical risk assessment methods were not instrumental, there was no instrumental assessment of work activities, movement analysis (kinematics, kinetics, sEMG, IMUs, hand-held dynamometers and grip force sensors) was not in the work activities or instrumental approaches were not used. A total of 30 articles were finally included in this review.
The international standards 11228-1, 2 and 3, 11226, TR 12295 and TR 12296 [77][78][79][80][81][82] accept and list the methods able to detect occupational physical risk factors and to evaluate the usefulness of ergonomic interventions. These methods consider several indices to be measured, from motion amplitude and frequency to exerted force [2].
The strength of these traditional approaches for biomechanical risk classification which can be used in a wide range of professional conditions, is based on their inexpensiveness and non-invasiveness. On the other hand, these methods have some weaknesses, mainly due by their observational nature and subjectivity related to the practitioner's expertise [1][2][3][4]. In most cases the worker's behavior is evaluated on pro-forma sheets either while observing in the field or replaying videos, an approach considered inaccurate, imprecise and time consuming [2,83,84]. Furthermore, the scientific literature highlights equations and parameter restrictions, with insufficient accuracy, unrepeatability and unreliability [4,85,86]. These concerns are mainly due to the necessity of assignment of scores to each risk determinant, such as upper and lower limb joint angles and range of motions, loads displacement, forces, work cycles, frequency of actions, forces and recovery times [85,[87][88][89][90][91][92][93][94].
Accurate and precise results could be better achieved by means of modern measuring devices, facilitating experts' diagnostics [2]. In recent years, instrument-based techniques designed based on current technological advances and performing direct measurements by using sensors attached to workers' bodies have been developed (see Table 1) and are now able to capture some or all of the parameters needed in the computation of the risk level. These instrumental and quantitative sensor-based tools might greatly increase the accuracy of these methods in ways that were not previously available [25,69,70,95] and widely reduce the time that an expert needs to carry out the same assessment manually. Finally, the use of automatic online tools would give a meaningful evaluation by gathering posture, kinematic, kinetic and muscular activity data in assessing WMSDs risk.

IMUs, Hand-Held Dynamometers and Grip Force Devices, sEMG Sensors: How They Are Made and Measure
Movement analysis systems allow, with a high accuracy and acquisition frequency, the quantification of motor functions, motor abilities, pathological conditions, compensatory motor strategies and improvements due to rehabilitation treatments and ergonomic interventions. However, these systems can be easily used only within the laboratory and more difficult in the field. This difficulty has led to the development, in the last decade, of accurate and reliable wearable human body sensor-based tools for easy human motion analysis directly in the workplace. The main factor allowing the abovementioned use has been, without doubt, the miniaturization of devices which has allowed huge benefits over traditional approaches. Other factors are wireless connectivity, light weight, small-size, low power consumption, portability, low-cost, comfort, and the possibility to monitor subjects remotely and to provide feedback to the end-user [19,[96][97][98][99][100][101][102].
Among wearable human body sensors, inertial measurement units (IMUs), dynamometers and surface electromyography (sEMG) sensors (see Figure 2) allow a detailed estimation (compared to traditional observational methods) of kinematics, kinetics and muscle behaviors without interfering with the typical movements performed by workers in the workplace [103,104].

IMUs, Hand-Held Dynamometers and Grip Force Devices, sEMG Sensors: How They Are Made and Measure
Movement analysis systems allow, with a high accuracy and acquisition frequency, the quantification of motor functions, motor abilities, pathological conditions, compensatory motor strategies and improvements due to rehabilitation treatments and ergonomic interventions. However, these systems can be easily used only within the laboratory and more difficult in the field. This difficulty has led to the development, in the last decade, of accurate and reliable wearable human body sensor-based tools for easy human motion analysis directly in the workplace. The main factor allowing the abovementioned use has been, without doubt, the miniaturization of devices which has allowed huge benefits over traditional approaches. Other factors are wireless connectivity, light weight, small-size, low power consumption, portability, low-cost, comfort, and the possibility to monitor subjects remotely and to provide feedback to the end-user [19,[96][97][98][99][100][101][102].
Among wearable human body sensors, inertial measurement units (IMUs), dynamometers and surface electromyography (sEMG) sensors (see Figure 2) allow a detailed estimation (compared to traditional observational methods) of kinematics, kinetics and muscle behaviors without interfering with the typical movements performed by workers in the workplace [103,104].
In experimental settings, IMUs, dynamometers and sEMG sensors are placed and fixed on the appropriate body segments to measure joint angles, forces and muscle behaviors, respectively. All the sensors are commonly synchronized for data alignment in time [105]. Connection is always performed by implementing one of two wireless protocols: Wi-Fi or Bluetooth. The former has an increased power consumption, but a greater transmission speed and distance with respect to the latter. The sample frequency of these sensors varies between 50 and 1000 Hz while the minimum number of bit is 12 [106].

IMUs
IMUs allow the measure of orientation, position, velocity and accelerations of each investigated segment and whole body posture. The term "inertial" comes from the fact that these sensors use the inertia principle: the acceleration can be related to the resistance to move (inertia) of a free mass accelerated by an external force or torque. In the literature, the type of sensors used ranges from uniaxial to triaxial accelerometers, gyroscopes and magnetic sensors [99]. Usually three orthogonal accelerometers and three orthogonal gyroscopes are embedded within the probe to measure linear acceleration and angular velocity, respectively, along three orthogonal axes. Angular displacements In experimental settings, IMUs, dynamometers and sEMG sensors are placed and fixed on the appropriate body segments to measure joint angles, forces and muscle behaviors, respectively. All the sensors are commonly synchronized for data alignment in time [105]. Connection is always performed by implementing one of two wireless protocols: Wi-Fi or Bluetooth. The former has an increased power consumption, but a greater transmission speed and distance with respect to the latter. The sample frequency of these sensors varies between 50 and 1000 Hz while the minimum number of bit is 12 [106].

IMUs
IMUs allow the measure of orientation, position, velocity and accelerations of each investigated segment and whole body posture. The term "inertial" comes from the fact that these sensors use the inertia principle: the acceleration can be related to the resistance to move (inertia) of a free mass accelerated by an external force or torque. In the literature, the type of sensors used ranges from uniaxial to triaxial accelerometers, gyroscopes and magnetic sensors [99]. Usually three orthogonal accelerometers and three orthogonal gyroscopes are embedded within the probe to measure linear acceleration and angular velocity, respectively, along three orthogonal axes. Angular displacements are obtained from numerical integration of the angular velocity while linear velocity and displacement are estimated from first and second numerical integration of linear accelerations. IMUs can also embed tri-axial magnetic sensors although their use turns out to be more critical in the workplace in presence of electromagnetic fields.

Hand-Held Dynamometers and Grip Force Sensors
Hand-held dynamometers, already described at the beginning of the previous century [107], are simple devices placed between a fixed place and the subject's body part to assess the isometric muscle (or muscle group) strength relevant as outcome measurements in studies evaluating changes in the functional status of joints, lower and upper extremities and trunk [108][109][110][111][112][113][114]. These devices are considered highly reliable [115], easy to use, portable, inexpensive and compacts if compared with isokinetic systems [112]. Besides hand-held dynamometers, forces are also recorded by using superior grip dynamometers although they can only be used for given hand sizes. Other critical issues related to grip dynamometers are the inability to measure forces from multiple fingers simultaneously, difficulty of use, low accuracy of grip force measurements and the difficulty to design appropriate handle shapes [116][117][118][119][120]. The measure of the grip force is also provided by instrumented gloves (i.e., equipped by force sensitive resistors) or by force sensor mats applied to handles [73,[121][122][123][124][125][126]. Instrumented gloves remove the need for the handle to be instrumented but disturb grasp interaction [127]. Furthermore, force sensor mats embedded within gloves acquire only normal forces, require calibration and may shift during measurements. In order to overcome these limitations multi-dimensional grip dynamometers have been developed to adapt to a wide variety of handle sizes and geometries allowing a continuous measure of fatigue and forces [119,127,128]. Recently dynamometers able to measure both angle and force with high levels of sensitivity and inter-examiner reliability have been developed [129][130][131][132]. Finally, it is interesting to also report haptic tools which consist of physical bendable strips allowing users to manipulate and apply deformations to digital surfaces and to move and rotate virtual objects. Such device allows a continuous, free hand contact on a developable strip bent allowing to industrial designers and stylists to perform an effective assessment of the aesthetic quality of the shape of new products and also its modification, directly on the digital prototype without the need to construct a physical prototype [133][134][135].

Direct Instrumental Evaluations
In this section of the results we report papers whose aims were to propose instrumental-based tools for biomechanical risk classification without using measured/calculated indices as input to standardized methods. Instrumental approaches based on wearable sensors have been used to classify lifting tasks into low and high risk categories. In a very recent study [173], IMUs and sEMG sensors have been used to monitor trunk inclination and trapezius and erector spinae muscle activity, respectively, during the execution of several types of lifting tasks with different weights, horizontal distance and technique executed by male office workers. The method proposed in this study allows, with an acceptable accuracy, the automatic identification of the risk levels associated with the lifting activities. Indeed, the lifting tasks were characterised by a feature vector composed of either the 90th, 95th or 99th percentile of sEMG activity level and trunk inclinations during the task. Linear Discriminant Analysis and a subject-specific threshold scheme were applied and lifting tasks were classified. The authors of this study highlighted how the strength of this study lies on its objective instrumental approach based on subject-specific thresholds and on the possibility to complement the current standardized approaches usually used to detect biomechanical hazardous.
Another recent kinematic-based lifting tool has been designed by monitoring several lifting tasks with growing lifting index (LI) computed by the revised NIOSH lifting equation [174]. Kinematic data allowed the calculation of a mechanical lifting energy consumption (LEC) index which proved to be significantly growing with the LI, discriminating all the risk condition pairs and well correlating with compression and shear forces that determine injuries at the L5-S1 joint. The findings of this study suggest a potential use of IMUs-based lifting tools in indoor and outdoor work environments for risk estimation.
Furthermore, muscle coactivation has deeply been investigated [66] because it, being a neuromuscular pattern needed to stabilize the trunk [57], represents one of the causal pathways for WLBDs. The behavior of the cervical and lumbar spine has also been investigated in complex multiplanar dynamic motions including lifting and pushing [175,176]. sEMG has been used to develop a sEMG-based multi-muscle coactivation index that resulted usable to continuously assess the neuromuscular effort and significantly sensitive to several factors. In particular the higher the speed, complexity of the motion and head control are, the higher the coactivation index value is. Also, in this case this simple approach has been proposed to be used for ergonomic assessments.
Another tool developed to calculate the simultaneous activation of trunk muscles is the time-varying multi-muscle co-activation index (TMCi) which includes a sigmoid-weighting factor dependent on relative differences between muscles that do not rely on a priori definitions of agonist or antagonist behavior [156]. This index was evaluated during the execution of lifting task in controlled conditions considering trunk muscles [156]. It has been shown that heavier lifting conditions resulted in higher TMCi values and that significant correlations exist between the TMCi and other agonist-antagonist methods. The same experimental setup used for LEC calculation [174] allowed to understand that also simple sEMG parameter values (i.e., ARV and max), besides TMCi, increased under heavier lifting conditions [177].
Moreover, sEMG data were also used in a study to implement tools based on an artificial neural network [178]. In this study sEMG features (i.e., max, ARV, mean and median frequencies) were evaluated during the execution of lifting tasks starting from the trunk muscles. Then, these features were used as input variables of artificial neural network for the prediction of WLBDs. The results show that sEMG time and frequency features are significantly related to lifting index for specific trunk muscles. Furthermore, the findings show that a tool based on these machine-learning techniques and sEMG feature, choosing a proper combinations of input features and a right network architecture, can lead to an improved biomechanical risk classification. Moreover, the authors concluded that the possibility to implement the integrated approach on electronic smart devices (smartphones, phablets, tablets and smartwatches) would allow a simplified analysis of biomechanical risk at workplace.
Manual lifting has also been assessed by using muscle fatigue estimation. In a recent review regarding this issue a list of methods was given, though the authors concluded that there are still many gaps to be filled and further studies are needed to find better fatigue indices and improved techniques [179].
Besides lifting activities, wearable sensors have been used for direct instrumental evaluations in handling of low loads during high frequency activities. For instance, local myoelectric manifestation of muscle fatigue estimation [160,161,180] has been estimated in several conditions to investigate several groups of workers. In particular it was used to investigate: (i) the biomechanical exposure of younger and older groups [181,182]; (ii) changes in fatigability in jobs with and without pause [166]; (iii) upper limb and trunk muscles in simulations of assembly tasks and in different light levels of repetitive work [162][163][164][165]167,[182][183][184][185]. Further experiments have been designed and performed in a recent study of Ranavolo and colleagues [58] in which the association between the local elbow flexor muscles fatigue and physical demand has been investigated. The findings of this study showed that local muscle fatigue estimated by using miniaturized sEMG sensors placed on the brachioradialis muscle is a promising index because of its sensitivity to the risk classes. Kinematic and sEMG assessments have also been performed in biomechanical evaluation of supermarket cashiers before and after a redesign of the checkout counter, in analysis of post office employees' workstations and in manual handling on a supermarket greengrocery shelf [186][187][188].
Finally, the usefulness of wearable sensors has also been investigated in many work tasks which require intensive and repetitive production of forces on the upper extremities in manipulating external loads, wrists, palms, fingers and tendons [189][190][191]. In these cases, the role of wearable sensors, in most cases hand-held dynamometers devices, is to measure the normal and shear forces created between fingers and handles to assess muscle integrity and to determine the level of any strength deficits [192,193] associated to clinical physical examination tests (i.e., the diagnosis of shoulder pain [194]). A wearable, unobtrusive, wireless and accurate system (Activity Tracking with Body Area Network) has been designed to operate autonomously to quantitatively measure the postures and body motions of workers [2]. This system is meant to be used by workers to autonomously monitor themselves on actual job sites over long periods of time. Different working processes in a wood workshop have been evaluated by using three accelerometers and two microphones and by correlating the worker's motion and frequency and intensity of sounds [195]. IMUs were also used in several work activities such as car assembly, hammering, screwing and drilling [196,197]. In construction activities IMUs [198][199][200] and sEMG has been used to monitor lifting and holding loads activities to detect potential sources of WMSDs at neck [201] lower back levels [202].

Risk Assessment in the Context of Rating of Standard Methods
As done for direct instrumental evaluations, in this section tentative ratings of standard methods using wearable technologies are analyzed. An innovative "on-body wireless sensors network"-based approach for real-time ergonomic assessment in industrial manufacturing has been proposed by Vignais and colleagues [84]. The sensor network was composed by IMUs and goniometers and the body posture (joint angles) was assessed by using a ten rigid segment, twenty degrees of freedom biomechanical model. IMUs were placed bilaterally on the upper arm and forearms, on the head, trunk (on the chest) and pelvis (on the sacrum). Goniometers were placed on the hands and forearms to measure wrist motions. Angle values were used as input within the Rapid Upper Limb Assessment (RULA) method, whose global and local scores were continuously computed by a mobile processing unit (a standard laptop) and fed back to the user via a see-through head-mounted display.
Moreover, a real-time body sensors network composed by IMUs and sEMG sensors has also been used in real-time to monitor workers by measuring muscular efforts and postures (upper limbs have been modeled as a 7-DoF kinematic chain) for WMSD prevention according to the RULA index and the Strain Index (SI) [3]. An interesting index considered for this tool is the percentage of time spent in every RULA score range by every worker, considering the whole experiment duration. The accuracy, expressed as the number of correct assessments (with respect to those performed by two human evaluators) of the system and the number of cycles, was 95% for RULA and 45% for SI, indicating that the body sensor network is able to give a RULA score estimation congruent to the one given by the human evaluators. As far as the SI score is concerned, the system gives a score congruent to the evaluators' evaluation in almost the 50% of the cases. sEMG has also been used for complementing the RULA scoring system [203] and as an alternative to the visual inspection according to the BORG scale. It is in fact demonstrated that the two assessments are strongly correlated [87]. An example of the latter application has been studied by Cabeças [204] where sEMG was used as an alternative to observational methods in computing the SI score. The authors concluded that, once appropriate trigger levels for the muscular activation are defined, sEMG is a valid alternative to visual inspection in SI computation. This is true in particular when efforts are not clearly associated to hand/wrist movements and when non-cyclical high-frequency activities are assessed.
In these 13 studies, the most investigated working tasks are lifting tasks [156,173,174,177,178] and manual handlings of low loads at high frequency [58,186,188]. As regards lifting tasks, all the results of these studies show that kinematic [173,174] and sEMG-based [156,173,177,178] indices significantly grow with the LI discriminating all the risk condition pairs. Moreover, when a correlation analysis has been performed, results show a good correlation between kinematic and sEMG-based indices and compression and shear forces [174,177]. Moreover, only one study [178] used machine-learning techniques and sEMG features. Results of these studies led to an improved biomechanical risk classification. As regard the studies regarding the manual handlings at low loads at high frequency, the results show that sEMG indices were sensitive to the risk classes [58,186].
Many of these studies investigated specific work activities [3,84,87,186,188,204]: handling of low loads at high frequency on a redesigned checkout counter [189], laparoendoscopic single-site surgery [203], real-life operations of super-market cashiers [3], manual tasks in an industrial environment [84], sawmill work [87], manual handling on a supermarket greengrocery shelf [186], cleaning activities [204]. The results of these show that kinematic [3,84,87,186,188] and sEMG [3,87,186,188,204] data could be used in the risk assessment in work activities.

Discussion
In the attempt to reduce the risk of work-related musculoskeletal disorders several methods have been developed, accepted by the international literature and used in the workplace. In the last years, the most innovative wearable technologies and electronic smart devices, without interfering with the work activities performed by workers, have been introduced to improve the biomechanical risk assessment adapting it to all the work conditions and overcoming the limits of the current standardized methods. Indeed, these devices allow the estimation of biomechanical risk in real-time providing a direct feedback to the end-user who would be constantly monitored directly while at work.
In this review, we report on recent implementations of wearable sensors for quantitative instrumental-based biomechanical risk assessments in the prevention of WMSDs. Their use is desirable also in view of the concerns expressed within a recent article [206] regarding technical ISO standards on ergonomics and physical workloads. Muscle fatigue with a decrease in the mean power frequency and an increase in root mean square amplitudes was found both during continuous work and work with pause activities. The muscle fatigue was less pronounced when pause activities were introduced into the work. Fatigue patterns were lower during the second hour, indicating adaptation to the work task and work pace. The ratings of perceived exertion and discomfort were similar during work with and without pauses and were higher during the second hour of work Trunk extensors and flexors muscles sEMG activities of the trunk muscles and forces at the L5-S1 joint to identify sEMG-based indices related to the risk level and to the forces at the L5-S1 joint sEMG indices were proved to be significantly growing with the LI, discriminating all the risk condition pairs and well correlating with compression and shear forces that determine injuries at the L5-S1 joint sEMG data, muscle coactivation and forces at the L5-S1 joint In this "discussion paper" the authors underlined, among the others, how: (i) the production of these standards differed substantially from evidence-based practical guidelines; (ii) it is not clear why the ISO subcommittee preferred one method of risk assessment over others; (iii) some statements in ISO 11228 series appear to be based on personal opinions and in contrast with scientific evidence from the literature; (iv) ISO standards are an effort by a self-identified committee of interested people to agree on "how something should be made" in order to facilitate exchange of goods, services, or other similar endeavors; (v) ISO standards should be used with caution. For these reasons instrumental-based tool will play an increasingly important role in both direct evaluations and in the rating of standard methods, also in consideration that several factors implying work-related musculoskeletal disorders interact at the same time. Therefore, it will be crucial to monitor all of them by using more than one method at the same time ensuring a more thorough evaluation of risk factors. On the other hand, a lot of attention must be paid because the use of more than one method can rapidly lead to unacceptably high costs for the practitioner, both from a time and money viewpoint [207,208].
In this context, the technologies accredited to be used are without doubt inertial measurement units (IMUs), instrumented gloves and surface electromyography (sEMG) sensors, although other new tools are appearing in research laboratories and the workplace. Among these, smart footwear-based wearable systems [209] will surely be useful because they will permit, by recording ground reaction forces through integrated tri-axial force sensors, an inverse dynamics analysis [210][211][212]. For their simplicity, vision-based tracking systems are also potentially useful for the rating of standardized methods as proposed for assessing the movements of workers within quick exposure check tools [83]. Wearable miniaturized sensors can monitor workers' motor behavior if individually placed on the body segments or embedded in elastic suits. The latter use is also the most probable because the research activity is working fast on the development of artificial muscles, materials able to reversibly contract, expand, and rotate due to an external stimulus [198,213,214]. These devices, that can be enriched by several material characteristics, textile layers, elastic components, diagonal and lateral seams and pneumatic mechanisms [215] are envisioned as actuators for silent, soft and compliant assistive devices [216] acting as force multiplier systems by helping workers to reduce their effort. These suits/devices can also embed miniaturized sensors which will also serve for their control through, for instance, effective feedforward anticipation mechanisms. Furthermore, numerous devices have been developed to support the trunk during dynamic lifting tasks. sEMG will allow the detection of the early preparatory muscle activities to classify muscle loading and to initiate appropriate device activation. It has been shown that preparatory muscle activity can be leveraged to identify the intent to lift a weight up to 100 ms prior to load-onset [217]. The reduction of the effort will also be guaranteed by highly adaptive production processes.
Although the use of new innovative technologies for biomechanical risk assessment is only at the beginning (see Table 1), the literature shows that these instrumental approaches could be used to classify lifting tasks into low and high risk categories. The reported studies used wearable sensors, such as inertial measurement units, dynamometers and surface electromyography sensors, for biomechanical risk assessment in different work activities: lifting tasks [58,65,66,173,176,177,201] manual tasks in an industrial environment [84,164,183], manual handling [174,186], work in supermarket [3,188], repetitive work [165,166], cleaning activities [204], surgery work [203].
The results of these studies show that the indices used for the instrumented-based approaches are proved to correlate with the variables that determine the injuries. Particularly, two of the most promising indices/approaches proposed in literature for a work activity such as manual lifting, are the multi-muscle coactivation index [65,66,156,177] and machine-learning techniques based on sEMG features [178] while for this work activity, the literature shows that the fatigue indices need further elaboration [179]. Probably the most critical factor in lifting activities is the frequency of the lifting action which at the current status cannot be determined by instrumental methods [177]. It is necessary to be able to parameterize the risk levels associated to it. On the other hand, frequency of the actions is taken into consideration in work activities concerning the repetitive movement when the assessment is performed by using fatigue estimation before and after the work activity [58]. In some studies, wearable technologies are used for the rating of the standard methods (see Section 3.4). In these studies, the observed discrepancies between the evaluators can be due to several factors related to both the human and the procedural sides [203]: artifacts in the sEMG signals, difficulties in performing his real maximal voluntary contraction for the muscles considered, and underestimation by the human investigators of the actual efforts exerted by the workers.
From a technological point of view IMUs do not suffer from such limitations but if a high number of units is required for whole-body bio-mechanical studies in ergonomics, a high data transfer time could be required with both the Wi-Fi and Bluetooth protocols. Furthermore, IMUs fail to precisely measure translational motion and suffer from drift. Finally, IMUs can fail in the presence of magnetic fields in the workplace if they have embedded magnetic sensors. As regards limitations associated to sEMG, crosstalk muscle signals, electrode-skin impedance, noise and problems related to the electrode location, size, configuration and distance are the main critical factors [218]. To optimize the sEMG measures it is essential to use reference books such as the "Atlas of Muscle Innervation Zones" [219]. For both IMUs and sEMG sensors the energy consumption and the consequent battery discharge do not seem to be problems anymore, thanks to the long life of the most recent batteries. Two main limitations are ascribable to dynamometers: forces are commonly measured in only one direction and the form factor of the handles is not characteristic of most handles encountered during everyday work activities [220][221][222].
Finally, the heterogeneity in experimental procedures of the articles included in this review such as in working tasks, body segments, muscles and instrumental-based indices, has not allowed a valid complex statistical analysis combining the results as in other systematic reviews. For this reason, we have avoided a statistical pooling and summarized the data narratively [222][223][224][225]. The lack of a meta-analysis can be considered a limitation of the study.

Conclusions
The analysis of the papers reported in this review sheds light on the fact that still too few researchers foresee the use of wearable technologies for biomechanical risk assessment although the requirement to obtain increasingly quantitative evaluations, the recent miniaturization process and the need to follow a constantly evolving manual handling scenario is prompting their use. Therefore, the use of new innovative technologies for biomechanical risk assessment is only at its initial stage, but the authors of this review believe that this process is unstoppable, as it is happening in all the other areas of medicine and beyond. Obviously, it will be necessary for any validation to follow evidence-based medicine/policy/legislation multistep scientific approaches by designing rigorous laboratory and epidemiologic studies, by replicating them by independent research groups and by systematically evaluating them through transparent review processes. We are however convinced that, even if such use should fail in ergonomic practice, the huge knowledge that will derive from its experimentation will allow the optimization of the current standardized methods or the developments of the new ones.
Author Contributions: R.A. performed electronic searches. R.A., T.V. and A.S. assessed titles and abstracts of the articles. R.A., F.D., T.V., A.S. and S.I. resolved disagreements among reviewers by scheduling dedicated consensus meetings. R.A., F.D., T.V., A.S. and S.I. wrote the text and did work revision.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.