Optical motion capture systems are based on optoelectronic stereophotogrammetry and measure kinematics of gait in three dimensions (47–50). These systems include marker-based and marker-less systems.

Force platforms are steel blocks equipped with strain gauges or piezoelectric transducers measure GRF and can be embedded in a walkway or in treadmills for continuous recordings of multiple gait cycles. The gait cycle results in a repetitive and unique GRF pattern with precisely timed events such as heel-contact and toe-off that can be quantitatively assessed (60). Additionally, center of pressure (CoP) can be measured continuously between the body and ground as an indicator of balance. Force platforms are generally expensive and require dedicated laboratory environments and skilled technical personnel to operate. However, they can be used in conjunction with motion capture and EMG systems to provide joint kinetics (moments, power, and forces applied by each joint when braking or propelling) making them useful for laboratory-based assessments of gait and balance in pwMS. Additionally, graphical representations of gait, known as “butterfly diagrams,” can be produced that represent the 2D envelop of the GRF vectors during a step and could have clinical utility (25).

In-floor force platforms display high test–retest reliability for gait (61) and balance (62–65) variables. The reliability of treadmill-based force platforms for simple gait variables (mean stride frequency, stride width, time and length, and double stance phase) is also high (ICC = 0.86–0.97); however, for more complex measures such as gait variability, the reliability is low to moderate (ICC = 0.22–0.44) (23). Significant differences also exist in the GRF patterns during treadmill walking compared to overground walking, so it is unclear whether treadmills are optimal for identifying pathological gait function in neurological diseases (66–68).

In pwMS, force platforms have been used to study changes in gait initiation, postural stability, and balance associated with therapeutic interventions and disease progression (46, 69–72). Notably, Orsnes et al. (73) examined the timing of heel-contact and toe-off events in pwMS treated with baclofen (an agent used to treat spasticity in pwMS) using treadmill-embedded force platforms. The investigators observed only minimal improvements in gait and balance with treatment. A more recent study employed treadmill platforms to study both walking and jogging in minimally disabled pwMS (mean EDSS = 1.8) (25). Compared to controls, patients displayed greater step time difference between left and right feet and increased step width during both walking and jogging, but with greater change during jogging. The authors also noted that variability in the location of the CoP throughout gait cycle correlated with EDSS cerebellar scores.

Portable balance boards provide an alternative to laboratory-based force platforms. These boards use four force transducers (one on each corner of the platform) from which the CoP position can be calculated using suitable software (26). Nintendo Wii Balance Board (Nintendo, Kyoto, Japan) is the most widely tested balance board due to its low cost, portability (weighing only 3.5 kg), and wide availability. Wii Balance Board is suitable for clinical, laboratory, and home testing and demonstrates good test–retest reliability (ICC = 0.66–0.94) and construct validity when benchmarked against laboratory-grade force platforms (ICC = 0.77–0.89) (26, 27).

Wii Balance Board has been used with custom software to study postural sway in pwMS (28). Compared to laboratory force plates, Wii tended to overestimate postural sway although the test–retest reliability of the Wii has been found to be high (84%) (26–28). Castelli et al. (28) were also able to discriminate pwMS who reported fallers vs non-fallers. A case study employing Wii Balance Board noted changes in balance recorded during an exercise intervention in a single participant who had a relapse in the 6-week intervention period (29). The authors suggested that balance changes could provide a means to predict relapse onset (29). Several trials using Wii Balance Board have been undertaken and have shown potential improvements in mobility balance and QoL in pwMS (74–77), indicating that physical programs using this low cost technology could be useful for patients’ physical therapy. Overall, the cost and weight advantages of Wii, together with its high reliability and validity, make it a useful tool for assessing balance in MS in the clinic and home. Further investigations are required to identify the most useful measures that can be obtained from the device for clinical monitoring.

Instrumented walkway mats are portable mats a few meters in length with sensors to identify foot contacts. GAITRite is the most commonly used instrumented walkway mat and can determine spatiotemporal measures of gait (including walking speed; step and stride lengths; base of support; step, stride, swing, stance, single support, and double support times; and toe in/out angle) with high sensitivity for detecting pathology-related changes (30, 32). Spatiotemporal outputs from GAITRite do not require skilled personnel for analysis and interpretation (78), facilitating its use in clinical settings. The GAITrite has been validated against highly advanced motion capture systems for spatiotemporal measures (33, 34) and has high test–retest reliability (ICC 0.82–0.98) (79, 80) for most gait variables in young and older healthy adults at preferred and fast walking speeds.

In pwMS, GAITrite measurement of gait variables, including time to complete, velocity, cadence and number of steps, velocity, swing time, and single support, have been shown to be comparable to the T25FW in detecting gait dysfunction (36) and correlate with cerebellar EDSS subscores (38). GAITRite is also sensitive to changes in gait in very early-stage MS in patients with minimal disability (35, 37). A key shortcoming of GAITRite for gait assessment in pwMS is the restriction of data capture to a few steps at a time. Therefore, GAITRite provides no information regarding longer-term variability on any gait measures (81), measures that have been suggested as an indicator of gait dysfunction in pwMS (14, 37).

Overall, non-wearable systems provide the most comprehensive measurements of gait kinematics available. These measurements are highly accurate, reliable, and sensitive to pathological changes, even early in the disease when clinical assessments lack sensitivity. However, these systems can be costly and are difficult to deploy in environments where everyday activities are performed (82). Low-cost marker-less optical motion capture systems such as Kinect, and portable balance boards such as Wii Balance Board, could overcome these problems, especially in clinical settings; however, as we discuss in the next section, the development of wearable technology could provide gait assessment in the community over longer time periods.

Pressure sensors are instrumented insoles placed or integrated into the shoe to measure changes in pressure between the foot and the ground. These sensors are comparable to the force platforms as they also measure the force from the ground applied to the foot, but unlike force platforms, they measure the force irrespective of its components in different directions (i.e., x-, y-, and z-axes) (39). Pressure sensors use plantar pressure measurements to calculate spatial-temporal gait variables, including phases of gait (e.g., stance time and swing time), and step time, length, and frequency (39, 83). There are a wide range of systems that use electromechanical sensors for plantar pressure analysis including capacitive, resistive, and piezoresistive sensors (39). When compressed, they calculate variations in applied load measuring proportional change in voltage (capacitive), conductance (resistive), or voltage (piezoresistive) (39). Arrays of sensors in configuration can measure plantar pressure in a matrix along the entire plantar surface.

The accuracy of discrete pressure sensor systems is comparable to optical motion capture (5.4% mean error) (40), external pressure calibration (ICC = 0.99), and when multiple insole pressure sensor systems are compared (ICC > 0.95) (39). In general, discrete and matrix pressure sensor insoles have good to excellent reliability for pressure measurements within and between trials (ICC = 0.80–0.99) (39, 84). However, as gait speed affects plantar pressure, it is recommended that gait speeds are controlled when collecting gait data with pressure sensors (84).

Three studies have used pressure sensor technology to study gait dysfunction in pwMS. One study used discrete pressure sensor insoles combined with mobile technology that included a hand held mobile device, to assess plantar pressure and step timing and observed greater plantar pressure in stance phase and greater variability in step timing in pwMS compared to controls (41). Two related studies assessed gait in early-stage MS (4) and changes in gait over the subsequent 12 months (52). Cross-sectionally, pwMS patients with pyramidal signs displayed increased double limb support and decreased walking speed and stride length compared to those with no pyramidal signs (4). Longitudinally, pwMS exhibited a decline in gait performance over 12 months in the absence of EDSS change (52). These results demonstrate that pressure sensors have the sensitivity to detect gait dysfunction in patients with no or minimal clinical disability.

Inertial sensors measure an object’s acceleration and can also be used to report velocity, orientation, and gravitational forces. Inertial sensors are the most widely used type of wearable systems for gait and balance analysis and have been validated in healthy volunteers and in groups with motor impairment (85–87). The most promising inertial sensors for 3D gait analysis consist of a combination of tri-axial accelerometer, tri-axial gyroscope, and tri-axial magnetometer. Tri-axial sensors can capture spatiotemporal (e.g., swing time and cadence) and 3D kinematic data including joint and segment angles. Similar to the pressure sensors, inertial sensors can be integrated into insoles making them highly suitable for gait analysis. However, they can also be attached to other parts of the body such as on a belt or the wrist as illustrated in Figure 1. Additionally, technology is being developed for inertial sensor data collection, storage and/or transmission with smart devices such as phones and watches (88–90).

Trunk- or shank-placed inertial sensors have been used to study gait dysfunction in pwMS, commonly during the TUG test (termed “instrumented TUG”) (14, 42, 44, 46, 91, 92). Spain et al. (91) reported increased sway acceleration during quiet stance with eyes closed and increased trunk motion during instrumented TUG in pwMS with normal walking speed. In a follow-up longitudinal study (14), the authors assessed changes in gait and balance over 18 months, demonstrating no worsening of balance and objective gait measures (sway and gait velocity, respectively), but differentiation of mild MS (average EDSS = 2.2), moderate MS (average EDSS = 4.3), and control groups based on gait velocity, trunk motion, sway range, and sway area. Variability in sway area, sway range, and trunk motion over time were significantly different between all three groups. Similarly, Solomon et al. (93) found that inertial sensor data differentiated pwMS and no clinical gait dysfunction from controls using measures of postural sway (mediolateral sway path length and mediolateral sway range). Importantly, inertial sensors during TUG appear to be quite reproducible (ICC > 0.85 for all trunk and shank recordings from pwMS tested over two sessions), and some variables (stride velocity, cadence, and cycle time) correlate significantly with EDSS and number of recent falls (92).

The great advantage of wearable sensors is the ability to measure gait in an individual patient’s everyday environment for extended periods of time. These systems now employ small wireless sensors that can remotely send signals to the laboratory or clinic. Connectivity between wearable systems and ubiquitous smart phones and watches could further improve the usability of these devices. Importantly, the cost of wearable sensors is generally lower than non-wearable systems making analyses on large numbers of patients feasible. Finally, wearable systems actively engage the patient in both assessment and rehabilitation and could reduce clinic visits by providing more real time information to the patient and treating clinician (94).

Wearable systems also have certain disadvantages. First, wearable sensors can generally measure a smaller number of gait variables than non-wearable laboratory systems. Therefore, early studies of wearables should involve benchmarking and validation against these more comprehensive systems. Second, the placement of the sensors on body parts could hinder daily activities, though this could be improved with integration of sensors into clothing, smartphones, and watches. Third, algorithms used to measure speed and distance with wearable systems can lead to amplification of measurement error (95). Indeed, the algorithms required to calculate gait variables, which in some cases require technical personnel to implement, are currently a barrier to clinical application. However, algorithm development is an active area of research and clinician and patient interfaces continue to improve (96). Finally, the use of wearable sensors by patients themselves in uncontrolled everyday environments can make them more susceptible to signal noise (e.g., magnetic or vibration interference), leading to incorrect data and inadequate durations of recording when out of the clinic (97).