The presented results confirm the hypothesis that inertial measurement systems provide objective and quantifiable measures of different gait deviations in neurological patients. We established a setup of inertial sensors and pedobarographic measurements to collect the objective balance and kinematic motion data in an outpatient setting without the need for a clinical gait analysis laboratory. A laboratory-free motion analysis seems important, as the environment was previously found to significantly influence gait performance when comparing neurological patients under laboratory and real-world conditions [6]. Although the outpatient setting of the present study does not fully correspond to the real-world situation (which would require measurements to be obtained at home or outside in daily life), walking measurements obtained in a standardized in-door (large corridor) environment provided nearly unrestricted measurement conditions compared to those confined to a pure laboratory setting. Moreover, the wearable sensors used in our study proved to be easily applicable and easy to wear, not restricting the natural movement capacity of subjects in any way.

Comparison of different patient groups to a reference group of healthy controls in our study revealed different patterns of motion and balance parameter alterations across different gait disorders: Regarding motion parameters while walking, NPH patients showed significant differences in nearly all walking parameters compared to healthy controls, whereas fewer of the walking parameters differed in the lumbar stenosis group (affecting shoulder, sagittal hip, and transversal ankle ROM, as well as maximum ankle plantarflexion), and only some parameters (transverse hip, and transversal/ plantarflexion ankle ROM) being altered in the CM group. On the other hand, balance parameters showed the strongest deviations compared to healthy controls in the CM group (regarding AP movement, passed COP distance, and by trend lateral movement), followed by LST patients (regarding AP movement and passed COP distance), and least in the NPH group (affecting only AP movement) (please see Table 1). Pedobarographic measurements were furthermore found to be sensitive for different conditions (eyes opened or closed), showing significant effects on all balance parameters analysed especially in CM patients. Our findings of differently altered balance metrics across patient groups comply well with the clinical observation of sensory ataxia to be mainly expected in CM patients (due to involvement of the dorsal columns of the spinal cord), but not typically in NPH patients.

Overall, NPH, CM and LST patients showed distinct patterns of movement and balance parameter alterations compared to controls. As there are only few studies which actually used inertial measurement systems to compare different pathologies, our study extents current knowledge, in that it showed distinguishable patterns of movement and balance parameter alterations across different gait disorders, providing a trajectory for using inertial measurement systems and pedobarography for the objective and quantifiable discrimination of different pathologies. Advancing the application of such technologies may aid to increase diagnostic accuracy and to economize diagnostic procedures.

The finding of a reduced ankle ROM in the transversal plane in our NPH patients may appear somewhat counterintuitive considering the clinically frequently observed outward rotation of the feet in NPH patients [34,35,36], but complies well with previous studies, in which restricted ankle movement was also reported in NPH patients [34]. By showing the overall rotation of the feet, we are aware that the outward placement of the feet is with respect to the line of progression in the transversal plane, and may arise from different levels (e.g. pelvis, hips, and ankles).

Arm swing has been described to be only mildly impaired in NPH patients, but has not been further analyzed in detail previously [34]. Our results show however a significantly decreased shoulder ROM in NPH patients compared to healthy controls (during one gait cycle by approximately 13°, Table 3), which may indicate a stronger hypokinetic impairment of the upper extremities in NPH patients than commonly noticed. The reduced hip ROM in the sagittal plane (during one gait cycle by approximately 10°) in our NPH patient group complies well with previous findings reported in the literature. Likewise, the reduced knee ROM or maximum knee flexion we found in NPH patients has been observed as well by other groups [34,35,36]. Both parameters were significantly decreased in NPH patients compared to CM patients, but not compared to LST patients. While it cannot be excluded, that the observed decreases in the ROM of different joints might be related to age effects rather than to specific gait disorders in our study, Stolz et al., previously described a decrease in ROM of the hip and knee in NPH patients, even compared to an age-matched reference group [34], which supports the results of the present study.

Kinematic movement parameter analyses in LST patients have repeatedly shown alterations particularly of lumbar and pelvic movement, with measurable impact of lower limb [26] and lower back pain [37] on movement patterns, and with objective and quantifiable measures of improvement after surgical intervention [28, 29]. One recent study [38] using optoelectronic techniques in a specialized laboratory setting described limited internal/external pelvic rotation and craniocaudal movement, limited hip extension and abduction/adduction, as well as limited ankle plantar flexion in this patient group. Our study confirms these previous findings in that the ROM of the hip and ankle were also restricted in our LST patients, while adding the observation of an additionally impaired ROM of the shoulders, which has not been reported before. In addition, we found balance metrics to be partially impaired in this patient group.

Concerning CM, kinematic movement data is also limited. In the literature, previous studies found significant differences in the knee ROM of CM patients compared to age-matched controls [21, 39,40,41], while another study found no significant impairment in the knee ROM during the stance phase [22], with the latter observation complying with our findings. Likewise, differing results have also been reported about the ROM of the ankle, which is increased in the sagittal plane in some studies [22, 39], while found to be decreased in another study [21], with the latter being as well supported by findings of our study. Such differences may in part relate to differences in the severity of clinical impairment among CM patients, as subclinical alterations of knee and plantarflexion have been reported in subclinical CM patients, as well as altered knee joint movement in severely affected CM patients (18), with an imbalance between agonist and antagonist muscles having been suggested as causative. CM patients previously showed aberrant sagittal alignment (with a larger anterior pelvic tilt and lumbar lordosis, but a lesser cervical lordosis and head flexion), impacting as well on biomechanics of the lower extremities (39). A previous study showed postoperative improvement in CM patients regarding balance, while movement parameters were not completely normalized (20). In another study, CM patients showed preoperatively greater range of motion of the ankle, the pelvis and the lumbar spine, but less ROM of the hip when compared to controls. Postoperatively, these patients showed increased knee and hip ROM, but lesser of the pelvis, the lumbar and cervical spine ROM (23), which may reflect regained postural stability after therapeutic intervention.

Different study results might relate to differences in the measurement setup and data processing or might be grounded in the study design. Accordingly, Kuruvithadam et al. [6] showed, that environmental factors and measurement conditions may significantly alter motion patterns. Walking speed for example affects walking patterns [42], which is also visible in our results, as e.g. hip ROM significantly increased during fast walking compared to normal walking. On the other hand, the knee ROM is not significantly affected by the walking speed, supporting the hypothesis of a higher influence of walking speed on the hip than on knee motion. Malone et al. analysed CM patients in comparison to age-matched healthy participants walking at comfortable and matched walking speeds [21]. They found significantly reduced hip ROM in the sagittal plane only while walking at a comfortable speed, whereas a significantly reduced knee ROM in the sagittal plane was found for both walking speed conditions [21]. Therefore, specific gait disturbances may only manifest depending on gait velocity. While we did not measure gait velocity directly, we were able to identify the indirect effects of gait velocity on different metrics by comparing movement parameters obtained during normal and fast walking. Analysing different walking conditions seems essential given such findings, as it may increase the sensitivity for detecting specific gait deviations. Nevertheless, significant interaction effects concerning diagnosis groups and different walking tasks were only for the parameter force distribution (AP) during static standing across different conditions, which might relate to the lack of statistical power of the present study.

We expected the dual-task condition during walking to be indicative of cognitive impairment accompanying gait disorders. When comparing NPH to CM and LSK patients, we thus expected gait performance in NPH patients to be significantly more impaired during the dual task condition. As opposed to our expectations, however, current results showed no interaction effect in this regard, which might relate to the small sample size, but which we believe should be readdressed in future studies with larger samples.

While many instrumented gait analysis studies in patients with neurological gait disorders focuse on spatiotemporal metrics as such as walking speed, stride length, stride width, or cycle variables, extending movement analyses by kinematic movement parameters with the inclusion of the upper extremities may add to a better understanding of the complex pathophysiology of human locomotion. Integrating these findings may aid in more precisely phenotyping different movement disorders and might help to identify additional treatment targets for physical rehabilitation. Future studies should investigate larger patient samples not only cross-sectionally but also longitudinally, and instrumented motion and balance parameters may well be increasingly subjected to data-driven analyses, potentially enabling automated classification algorithms soon.

Although a multitude of different movement and balance parameters were analysed in the present study, the small and heterogenous sample size has to be regarded as a major limitation. Based on the small number of participants, we are well aware of the exploratory character of this study, whose findings need to be validated in further investigations. Furthermore, age differences across groups may be regarded as a further shortcoming, as we cannot exclude that some of the measured differences could be confounded by age effects, although normal gait characteristics of young and healthy subjects are commonly referred to in the literature. However, becauseof disease-immanent age differences of various gait disorders (per se limiting direct comparability of different patient groups with each other), we rather chose a common normative control group of young healthy subjects as “a common reference frame”, rather than choosing different control groups for each gait disorder. Moreover, the recruitment of “healthy elderly subjects” may not necessarily exclude the existence of some underlying, yet unrecognized pathology, which would rather distort the “healthy” reference values. Potentially age-related effects in our sample may even be smaller than suspected, as a study by Renggli et al. [14] showed differences between groups with a mean age difference of 50 years to be larger in a real-world environment, while age groups differed in a non-real-world environment only in stride and gait velocity. Moreover, age differences were larger than in our study so age effects might be further limited here.

As inertial sensors might be prone to various errors (signal drift, magnetic smog, movement of the sensors after calibration on the body segments), repeatability of measurements might be limited and will require further technical refinement and validation, although validity and reliability of inertial measurement unit-derived kinematics have been described as excellent for mean spatiotemporal parameters during walking [43] and with good to excellent agreement for all sagittal kinematic parameters when compared to optical motion capture systems [44]

Although the presented measurement setup is easily applicable, it may not be widely accessible to clinicians yet. Considering the complexity of human locomotion and the high dimensionality of parameters obtainable by instrumented gait and balance parameter measurements, it is necessary to further define those metrics most sensitive to neurological gait deviations, to economize data acquisition and minimize post-processing time. Furthermore, aspects of test–retest reliability, as well as measurement accuracy of the applied techniques, have to be addressed first, before they might be established in clinical routine.