1000/1000
Hot
Most Recent
The occurrence of peripheral neuropathy (PNP) is often observed in Parkinson's disease (PD) patients with a prevalence up to 55%, leading to more prominent functional deficits. Motor assessment with mobile health technologies allows high sensitivity and accuracy and is widely adopted in PD, but scarcely used for PNP assessments. This entry provides a comprehensive overview of the methodologies and the most relevant features to investigate PNP and PD motor deficits with wearables. Because of the lack of studies investigating motor impairments in this specific subset of PNP-PD patients, Pubmed, Scopus, and Web of Science electronic databases were used to summarize the state of the art on PNP motor assessment with wearable technology and compare it with the existing evidence on PD.
Parkinson’s disease (PD) is a chronic and progressive neurodegenerative disorder, clinically defined by the presence of resting tremor, rigidity, and bradykinesia [1]. These features are collectively referred to as motor symptoms and mostly related to loss of dopaminergic neurons in the pars compacta of midbrain substantia nigra. Alpha-synuclein-positive intracytoplasmatic inclusions, known as Lewy bodies, are the pathological hallmark of the disease [2]. As the disease progresses, motor disturbances represent considerable illness burdens. Deficits in balance and gait are common and disabling features that significantly increase the patient’s risk of falling [3] and the managing of daily living activities [4].
PD is also characterized by strong clinical and neuropathological evidence of systemic involvement. The presence of Lewy bodies in several other nervous structures, such as the nervous fibers in the skin, indicate that peripheral nervous system (PNS) involvement may be an intrinsic part in the PD pathological process [5][6]. Since the PNS is a target of alpha-synuclein deposition, it is plausible that intrinsic pathogenic features of PD may predispose to peripheral neuropathy (PNP).
PNP refers to any disorder of the PNS including single and multiple mononeuropathies, symmetrical involvement of nerves (polyneuropathies), or isolated involvement of sensory ganglia (ganglionopathies) [7]. It usually starts gradually and presents in the most common types a distal-proximal gradient, affecting first the feet and later the hands [8].
The occurrence of PNP in PD (PNP-PD) has been shown to be present in up to 55%, compared to 8% in the general population with comparable age [9][10][11]. Typical features of PNP include postural instability, muscle cramps, and numbness, of which the latter two are more prominent at distal part of the legs. As both PD and PNP pathologies are associated with these symptoms, the concurrence of peripheral involvement could be considered as an additional cause of motor deficits and general worsening in PD [12].
PNP can worsen the global functional mobility of patients, since neuromuscular factors (hip strength, ankle proprioception, and decreased peripheral sensation) have been linked to gait and balance difficulties [13]. It is, therefore, plausible to hypothesize that PD patients with PNP (PNP-PD) may develop more prominent gait and balance deficits and, consequently, be at risk of falling, injuries, and reduced quality of life [14].
Wearables are constituted of all mobile devices worn on the body (also called on-body sensors), such as inertial measurement units (IMUs), smartwatches, or Holter electrocardiogram monitors [15]. They provide objective and quantitative measures from controlled and unsupervised environments, allowing the development of accurate treatment plans and disease monitoring. In particular, data obtained from IMUs can successfully estimate spatial-temporal parameters and provide sensitive and objective information about motor deficits of various neurological pathologies, which nontechnological motor assessments often cannot identify. Mobility assessment with wearable health technologies are widely investigated in a variety of illnesses, particularly in PD, and allows high sensitivity, accuracy, and reproducibility [16]. However, these methodologies are scarcely studied and have yet to be explored in PNP [17], although a small number of previous works using wearable sensors have successfully demonstrated motor and physical activity characteristics in PNP compared to controls [18][19]. Since the presence of PNP has only recently been considered related to PD, we were interested in understanding whether PNP-PD patients showed specific motor deficits, which can be measured with the use of wearable health technology. For such purpose, a preliminary review of literature performed by the authors showed no studies evaluating the functional impact of PNP in PD on mobility using wearables. Identifying specific gait and balance patterns in this specific subset of PNP-PD patients could provide additional information about gait and balance problems, which can be used to monitor and stratify patients, optimize treatment, prevent falls, and increase quality of life.
Multiple wearable sensor types were used within the included articles to assess measures of gait and postural stability in PNP patients. Among the 24 included articles, the most commonly used inertial sensors included a tri-axial accelerometer and a tri-axial gyroscope (83.3% of the studies): LegSys™ and BalanSens™ (BioSensics), used, respectively, for gait and balance assessment; the Opal v1 (APDM) and the Physilog® (BioAGM) for balance assessment; the GaitMeter™ for gait assessment; and the mHT (mHealth Tecnologies) for both gait and balance assessment. Accelerometers only were used in two studies: PAMSys™ (BioSensics) and DynaPort Mini-Mod (McRoberts BV). One study used a gyroscope-based sensor (SwayStar device, Balance International Innovations GmbH) for balance assessment [20]. Sampling frequencies between 50 and 200 Hz were used to acquire the signals. The most commonly used sampling frequency was 100 Hz.
Several sensor placements and numbers of wearable sensors were used, depending on the task and on the type of assessment. Among the 16 included studies analyzing gait in PNP, four papers (25%) used one sensor, four studies (25%) analyzed gait with sensors on both shanks (two sensors), one paper (6.25%) used four sensors, and six studies (37.5%) assessed gait with five wearable sensors placed on thighs, shanks, and lower back. One study did not report sensor placement (6.25%).
Postural stability was assessed in 13 studies: Three studies (23%) used one sensor on the lower back, five studies (38.6%) used two sensors, and two studies (15.4%) used three sensors on both shanks and lower back. The remaining three studies (23%) utilized five sensors (Figure 1, Table 1).
Figure 1. Anatomical representation of sensor placement for gait and balance assessment in patients with polyneuropathy (PNP).
There is currently no consensus available on the optimum number and placement of sensors to measure PD symptoms. All reviews included that evaluated sensor number and placement showed that the majority of the studies used one sensor placed on the lower back (at lumbar vertebrae level L3, L4–L5, sacrum, or waist) or on the dominant lower limb (thigh, shank, ankle, or foot). Single sensors seemed sufficiently robust for all applications: For gait assessment at home, one sensor was used in 28% to 47% of the studies [21][22][23], while for gait evaluation in the laboratory it ranged from 44% to 69% [24][25]. Not surprisingly, for balance assessment the use of one sensor, and specifically on the lower back, was preferred in 77% to 100% of the studies included in the reviews [25][26][27]. Other most commonly used sensor placements for PD were on both wrists or lower limbs (in 30% of studies) or on lower back and both lower limbs (in 14% of studies) for the home assessment and at both lower limbs (8% of the studies) for laboratory assessment (Table 2).
We included 24 original full-text manuscripts: Eleven studies (45.8%) investigated gait, eight (33.4%) analyzed balance, and five (20.8%) evaluated both gait and balance in PNP patients.
Gait was assessed mainly during a straight walking task at preferred gait speed, with a distance varying from 7 to 50 m. In two studies patients were asked to perform a 90° turn during walking [28][29]. Several parameters were calculated from the signals acquired through the wearable sensors. The most commonly reported parameters computed from the filtered signals were spatiotemporal gait parameters: gait speed (m/s), stride and step length (m), stride and step time (sec), number of steps, double limb support time (%), and cadence (steps/min). Coefficient of variation (CV) of gait speed and stride length and time (%) was calculated in eight studies [28][29][30][31][32][33][34][35]. Gait speed initiation, number of steps, and total distance required to reach steady-state walking were studied in four papers [33][34][36][37]. Duration (%) and number of walking bouts were extracted in one study [18].
Clinical trials among the included papers did not show any statistically significant changes in the gait parameters when comparing pre- and post-intervention. Najafi [38] analyzed gait differences between intervention and control groups after plantar electrical stimulation in DPN patients and Schwenk et al. [32] evaluated gait after a new interactive training in CIPN subjects. Nevertheless, the effect size of these studies suggested the presence of a moderate to large improvement of cadence and gait speed post-treatment. In contrast, Caronni [39] compared the responsiveness to rehabilitation in a group of PNP patients and found a statistically significant difference in gait speed between groups (p = 0.001, Table 1). Spatiotemporal parameters were significantly different between PNP patients and healthy controls only in studies investigating gait under more challenging conditions. Kang et al. [31] described a statistically significant difference between DPN and healthy participants in the coefficient of variation of gait speed and stride length during dual-task gait. De Bruin et al. [40] found significant differences in speed, step length, and cadence when comparing DPN patients during dual-task walking on paved trajectories compared to single-task. Another study by Kang [41] showed improvement in stride velocity, stride length, and double limb support (%) during dual-task and fast walking, compared to single-task, after plantar mechanical stimulation. Differences from controls were found in step time, cadence, and gait speed but not in stride length in a study by Esser et al. [17], and gait speed was also 10% decreased in DPN group compared to controls in a study by Ling et al. [30]. Another important result was pointed out by Najafi et al. [33], who found differences in spatiotemporal parameters only during long distances, especially in gait variability and in double support time, when comparing DPN patients with controls. These differences were more pronounced during barefoot walking.
Balance and postural stability were investigated through numerous tasks. The most frequently used task in all 13 studies was the double leg stance performed in different conditions:
Other tools to assess postural stability were clinical tests such as the functional reach test [44]. Functional tests (to investigate functional mobility, addressing both gait and balance characteristics) were performed in three selected studies [39][41][44]. They applied the timed up-and-go (TUG) test. This test was split by Caronni et al. [39] into five subphases, and the duration of each phase was measured, as well as the total TUG test duration.
The included studies reported multiple outcomes of standing balance and postural stability that were calculated from the signals provided by the wearable sensors (Table 1). Of these outcomes, the most commonly reported measures included center of mass (COM) sway (cm2), defined as total sway (in seven studies, 53.8%), and related parameters (anterior-posterior (AP) and medio-lateral (ML) sway (cm)). These parameters were also reported in three studies analyzing gait to investigate balance control during walking and gait initiation [33][34][37]. In addition, ankle sway (deg2), hip sway (deg2), and COM sway area (m2) were calculated in six papers (46.1%). Center of gravity (COG) sway (cm2), COG AP, and COG ML (expressed in cm) were calculated in one paper [45]. Other parameters were root mean square (RMS, m/s2), trunk acceleration, and trunk jerk (m2/s3) [39][46]; postural coordination of upper and lower body (defined as the reciprocal coordination between hip and ankle motions) [36]; roll and pitch velocity (deg/sec) and roll and pitch angle (deg) [20]. Further parameters were local (in short time intervals, sec) and central (in long time intervals) control balance strategies [45], and cross-correlation function (CCF) of angular velocity to investigate the coordination of human movements [46].
A significant reduction in COM sway area (a parameter of postural sway) was shown by Schwenk et al. [32] and Grewal et al. [47] after an interactive sensor-based balance training and by Yalla et al. [44] after an intervention on postural stability with an ankle foot orthosis. These results were found during balance tasks with open eyes, while, interestingly, no significant reduction was found during closed-eyes condition. In contrast, changes of the parameters COM sway area and ML sway area were significant after a virtual reality intervention with eyes-closed and -open conditions [35].
In PD, a multiplicity of parameters derived from inertial sensors could be described. For the purpose of this review, parameters from the upper part of the body (upper limb) were not considered. The included reviews listed a series of most relevant spatiotemporal parameters representative of five domains (pace, variability, rhythm, asymmetry, and postural control), which included stride length, stride velocity, cadence, double support time [48][49], and turning velocity [50] followed by step time variability [25][48] and step height, reaction time, and gait cycle duration [51]. Frequency-based measures were dynamics in trunk movement during gait, turning and smoothness [52], harmonic ratio, amplitude, slope and width of dominant frequency, peak trunk horizontal velocity, and phase coordination index of gait cycle [25]. Number of steps, single versus multiple step response, turning duration, turn-to-sit duration, and sit-to-stand and stand-to-sit time- and amplitude-based measures were reported to be important features to determine gait impairment [51]. In more detail, PD patients have been shown to have slower gait, less foot clearance, smaller step lengths, lower turning velocity, lower cadence, and lower peak trunk rotation compared to controls [48][50]. Turning velocity, cadence, and peak trunk rotation were associated with disease progression [53]. Another important parameter in PD is gait variability, also referred to as unsteadiness and arrhythmicity of stepping [54]. Increased gait variability can be seen throughout the disease, and the magnitude of the variability tends to increase with disease severity [48].
Home assessment may have greater ecological validity and gives a true picture of the burden of disease [15]. Parameters that may be particularly relevant for this assessment type are walking bouts (total number of walking bouts, median number of steps per bout, bout duration), turns per hour during the day, duration of each turn, number of steps per turn, peak and average rotational turning rate, and variability of these measures throughout the day and week [22][23].
Regarding standing balance and postural stability, often used parameters were postural sway velocity, RMS accelerations, and jerk [27]. Parameters that may discriminate most effectively between PD and controls are sway area, sway velocity, jerk index, sway amplitude and range of acceleration signals (time domain), and frequency dispersion and centroidal frequency [26][48] (Table 2).
All these features are able to differentiate between PD and healthy controls (HC) at early stage [25][48], different PD stages [27], different medication states in advanced PD, and PD progression (in particular sway dispersion and sway velocity) [48]. Postural sway is also a good measure of balance control to be used as a primary outcome for interventions [48].
Table 1. Summary of the major characteristics of the research design, analyses, and outcomes for the studies on PNP that met the inclusion criteria.
ACC: accelerometer; AP: anterior-posterior; CIDN: chronic inflammatory demyelinating polyneuropathy; CIPN: chemotherapy-induced peripheral neuropathy; COG: center of gravity; COM: center of mass; DFU: diabetic foot ulcer; DM: diabetes mellitus; DPN: diabetic peripheral neuropathy; Freq: sample frequency; GYR: gyroscope; MAG: magnetometer; ML: medio-lateral; NeP-DPN: neuropathic pain diabetic neuropathy; PNP-LL: peripheral neuropathy of the lower limbs; TUG: timed up-and-go test.
Table 2. Summary of the major characteristics of the PD reviews that met the inclusion criteria.
REFERENCE |
REVIEW CHARACTERISTICS |
NUMBER OF STUDIES INVESTIGATING PD |
SAMPLE SIZE (H&Y Stage) |
SENSORS (Number and Type) |
EXTRACTED PARAMETERS |
Morgan et al., 2020 [21] |
Analysis of gait during home assessment |
65 papers |
Almost half of the studies used between 10 and 49 PD participants. 12 studies used fewer than 10 and 8 more than 100 participants. |
45.5% of the studies used 1 sensor at the lower back; 2 studies used 3 sensors at lower back and feet; 1 paper used 1 sensor on the chest, 1 used 1 sensor on the wrist. 2 papers do not discribe the position |
Features not specified. |
Ghislieri et al., 2019 [26] |
Analysis of standing balance |
14 papers |
From 10 to 58 PD patients (and one study with 104 patients) |
The 93% of studies used 1 sensors on the lower back. 1 study used 3 sensors: 1 on the lower back and 2 on lower limbs |
Jerk index, sway amplitude, range of acceleration signals, frequency dispersion and centroidal frequency. |
Rovini et al., 2018 [22] |
Analysis of gait during home assessment |
30 papers |
Ranging from 1 to 75 PD patients |
6 papers (28.2%) used 1 sensor: 4 on the waist and 2 on the lower back. 10 (33.3%) papers used 2 sensors: 5 on the wrists, 1 on the feet, 3 on the ankles, one on ankle and dominant leg. 6 studies used 3 sensors on the waist and feet. 2 papers used 5 sensors (on wrists, ankles and trunk; on shanks, wrists and sternum). The last 3 papers used more than 6 sensors. |
Average time and distance walked, cadence, gait speed, step length, swing time, double support time; stride time and stride time variability. Inter-trial variability, inter-subject variability; inter-task variability. Number of turns per hour, turn angle amplitude, turn duration, turn mean velocity, number of steps per turn, hourly frequency of turning, duration of each turn, number of steps per turn, peak and average rotational turning rate, jerk, variability of these measures throughout the day and week. |
Merola et al., 2018 [51] |
Analysis of gait and balance |
6 papers |
From 6 to 40 (and 2 studies with 190 and 139 PD patients) |
Not reported |
Gait: temporal (reaction time, gait cycle duration), spatial (step length, step height) and biomechanical (ankle torque, vertical landing force) variables, and gait strategies (i.e., number of steps, single versus multiple step response). Balance and postural instability: trajectory of the center of pressure (COP) and center of mass (COM) misplacement, trunk acceleration and postural sway |
Vienne et al., 2017 [24] |
General analysis of gait |
16 papers |
Not reported |
11 studies (68.7%) described the assessment of PD with 1 sensor at the lower back. one paper used one sensor at one ankle, one at one shank and one at one foot. One paper used 2 sensors (upper and lower back), and one paper utilized 3 sensors at lower back and shanks |
Features not specified. |
Rovini et al., 2017 [52] |
Analysis of wearable sensors on support of PD treatment and diagnosis |
80 papers |
From 5 to 47 (and 1 study of 75 PD patients) |
Not reported |
Statistical (e.g., mean, variance, skewness, kurtosis), frequency (e.g., energy, power spectral density, fundamental frequency), and spatiotemporal/kinematic (e.g., stride length, TUG time, stride velocity) features; step or stride segmentation. |
Godinho et al., 2016 [16] |
Mobile health technology characteristics |
76 papers |
Not reported |
Not reported |
ISway measures (jerk, RMS amplitude and mean velocity from the time-domain measures, and centroidal frequency); gait parameters with a high degree of accuracy; total number of walking bouts, the percent of time spent walking, the total number of steps, median walking bout duration, median number of steps, and median cadence per bout. Quality-related sensor derived measures included: frequency measures, regularity measures and the harmonic ratio. |
Del Din et al., 2016 [23] |
Analysis of gait during home assessment |
19 papers |
From 2 to 169 PD participants (and one study of 467 patients) |
9 studies (47.3%) used 1 sensor on lower back; 3 used 2 sensors on thighs; 2 papers used 2 sensors on feet; 1 on both shanks and 1 used 1 sensor on the chest; the other papers used more than 4 sensors. |
Number of walking bouts, walking duration, total number of steps, median number of steps per bout, bout duration, cadence, step and stride regularity, frequency domain measures (harmonic ratio, amplitude, slope and width of dominant frequency), step duration, step symmetry, acceleration range and dynamic stability |
Oung et al., 2015 [49] |
Assessment of motor disorders in PD |
Not reported |
Not reported |
Not reported |
Step frequency, stride length, entropy and arm swing |
Hubble et al., 2015 [27] |
Analysis of standing balance and walking stability |
26 papers |
From 5 to 67 PD patients |
20 studies (76.9%) used 1 sensor on the lower back (sacrum/L3/L4/L5); 2 studies used 2 sensors on the shanks; 2 studies used 1 sensor on sternum/chest; 1 study utilized one sensor on the wrist; and another one on the lateral side of the pelvis. |
Sway velocity (23% of studies), RMS accelerations (19% of studies) and jerk (19% of studies). Harmonic ratio (31% of studies) and stride time variability (27% of studies). |
Steins et al., 2014 [50] |
Assessment of functional activities with wearable devices |
6 papers |
Not reported |
Not reported |
Stride length, stride velocity, cadence, and turning velocity |
Maetzler et al., 2013 [25] |
Quantitative objective assessment of gait and balance |
16 papers |
Not reported |
Gait: 4 papers used one sensor on the lower back (44.4%). 2 papers utilized 1 sensor on the shank and 2 papers 2 sensors on both feet. 1 paper used 1 sensor on the forearm and two studies used more than 5 sensors. Balance: 5 papers used 1 sensor on lower back (100%). |
Gait: Phase coordination index of gait cycle; stride length; frequency-based measures of gait (harmonic ratio, amplitude, slope and width of dominant frequency); cadence, step time variability; peak trunk horizontal velocity, turning duration, turn-to-sit duration; time- and amplitude-based measures of sit-to-stand and stand-to-sit; peak trunk rotation velocity and rotation range of motion, turning velocity; Walk peak roll velocity, total turning duration, turn peak yaw and roll velocity. Balance: Velocity, jerk, acceleration, frequency-based measures; displacement, velocity; Peak trunk acceleration during anticipatory postural adjustments towards the stance leg; Hilbert-Huang transformation of postural parameters |
Horak et al., 2013 [48] |
Biomarkers of gait and balance |
Not reported |
Not reported |
Not reported |
Gait: Stride Time Variability, double support time, peak arm velocity, trunk rotation, gait velocity, cadence, stride length. Balance: Postural sway (area, velocity, frequency) and jerk. |