The testing was performed in one 60-min session at either the Vestibular Research Laboratory in the Department of Speech Pathology and Audiology at the University of South Alabama or at the University of South Alabama Neurology Clinic. In order to complete the experimental tasks, all participants were required to possess the strength and stamina to stand unsupported for the duration of 60-s intervals, in multiple test conditions. The test procedure (see below) required approximately 10 min to administer and both healthy subjects and patients had a positive attitude during these testing paradigms.

Exclusionary criteria for the control group included history of previous neurologic and otologic disease, and additional significant medical history. Exclusionary criteria for the PD group included history of previous otologic disease, other neurologic disorders, and additional significant medical history. All participants read and signed a statement of informed consent approved by the Institutional Review Board at the University of South Alabama.

Sixteen individuals grouped based on their history of diagnosed Parkinson's disease (PD) participated in the present study Three males and two females served as participants in the control group (n = 5). Individuals in the control group ranged in age from 48 to 69 years (61 yo; 59–68), and were negative for history of neurologic incident or diagnosis, balance disturbance, vertigo, and significant otologic pathology. The PD group (n = 11) consisted of seven males and four females, and ranged in age from 60 to 83 years (65 yo; 60.5–71). Time since PD diagnosis preceding experimental testing ranged from 1 to 34 years (4 years; 3–7.5). The two groups did not significantly differed in age (P > 0.1).

In order to assess the impact of dizziness, imbalance, or unsteadiness on daily life activities, all participants completed the Dizziness Handicap Inventory (DHI; Jacobson and Newman, 1990). The DHI questionnaire contains 25 items, identified in three subscales: Functional limitations, physical movement, and emotional well-being. Items in the functional limitations subscale assess the extent to which dizziness, imbalance, and unsteadiness limit participation in normal daily activities. The physical subscale assesses the impact of specific head and body movements in precipitating and exacerbating feelings of dizziness, imbalance, and unsteadiness. Items within the emotional subscale assess the impact of the dizziness, imbalance, or unsteadiness on the individual's emotional well-being, such as feelings of depression, social isolation, and the effect of the problem on personal relationships. Participants were required to respond verbally to each item using the forced-choice, closed-set response alternatives “yes,” “sometimes,” or “no.” Participants were instructed to respond to each item as it pertained only to dizziness, imbalance, or unsteadiness experienced, separating this from the effects of Parkinson's disease as much as possible. Scoring of the DHI was completed such that “yes” responses were assigned four points; “sometimes” responses were assigned two points; and “no” responses were assigned zero points. Therefore, the highest possible score on the DHI was 100 points, indicating significant handicapped related to dizziness, imbalance, or unsteadiness. Total scores on the DHI close to zero indicate little to no impact of dizziness, imbalance, and unsteadiness on daily life.

The modified Clinical Test for Sensory Interaction on Balance (mCTSIB; Shumway-Cook and Horak, 1986) was used to assess each participant's functional ability to maintain balance in four conditions. Participants were instructed to maintain a quiet stance with bare feet approximately shoulder width apart on: (1) a firm surface with eyes open; (2) a firm surface with eyes closed; (3) a compliant surface with eyes open; and (4) a compliant surface with eyes closed. Each condition was maintained for 60 s, or until a corrective step was made or fall was imminent. This duration of testing was chosen to improve the sensitivity in the dynamical analyses of the center of pressure. During the firm surface conditions, participants stood directly on the Balance Board. High-density foam (18″ × 24″ × 6″), marked for correct foot placement, was used in the two compliant surface conditions. The foam was of sufficient density to prevent the participant from sinking to the firm surface below, thereby eliminating reliable somatosensory cues. Slight modification of the foam, wherein a small overhang was added to the foam block, allowed for direct placement of the foam atop the Balance Board without movement during testing. Sensitivity of the Balance Board to force on each sensor was maintained during testing with the high-density foam. Conditions were tested in a standardized order: (1) hard floor and open eyes, (2) hard floor and closed eyes, (3) foam and open eyes and finally (4) foam and closed eyes. Patients were allowed to relax between testing. Special marks on the board were used to reduce the variability in foot positioning between patients or when patients needed a break between conditions.

Conditions in the mCTSIB selectively distort or eliminate the visual and somatosensory input used to maintain balance. Specifically, visual input was removed in the two eyes closed conditions (conditions 2 and 4), and somatosensory input was made unreliable in the two high-density foam conditions (conditions 3 and 4). Therefore, in condition 2, the participant was only able to rely on somatosensory and peripheral vestibular input to maintain balance, and in condition 3, the participant was only able to rely on visual and peripheral vestibular input to maintain balance. In condition 4, the visual input was removed (eyes closed) and the somatosensory input was distorted and unreliable (foam); therefore, the participant had only input from the peripheral vestibular apparatus to maintain balance and upright posture.

During the mCTSIB, measurements of center of pressure were made using a Nintendo Wii Balance Board. The Balance Board has a sensor on each of the four corners, which measure the force of each foot in the specific quadrants of the board. Information gathered from the sensors was sent to the interfaced computer and software, based on the BlueTooth Toolbox (http://forums.ni.com/ni/attachments/ni/170/265158/1/wiimote.zip). The values (expressed in Kg) from the four (left front, LF; right front, RF; left back, LF; right back, RB) sensors were captured at a sampling rate of 20 Hz and normalized by the weight of the patient. The weight of the participant (Kg) at each time was calculated according to the following equation: (1)Weight (Kg)=∑values of each sensor102

The time series generated by each sensor were recorded in a TMS file (labview format) and stored on the hard disk for off-line analyzes in matlab environment.

Specifically, four successive sequences of recording were run for the four paradigms previously described. Each recording sequence was started by the investigator and ended automatically at the end of the 60-s test interval. The investigator stopped the recording if a corrective step was made or fall was imminent.

All the analyses were performed off-line in Matlab environment.

The euclidian coordinates of the center of pressure were first normalized by: (2)sn (t)=s (t)−mean (S)std (S)  with S defined either by X or Y

In which Sn is the normalized signal, S is the original signals, mean is the average and std is the standard deviation of the original signals. Preliminary study showed this technique useful to reduce the variability between patients and causal to their differences in weight.

Polar coordinates were then calculated from the normalized Euclidian for orienting geometry to the position of center of pressure between [−π, +π] with a zero angle at the median of the frontal segment. The polar angle (or azimuth) is defined by: (3)α=atang2 (Y,X)

The azimuth was used for the analyses described below.

Preliminary principal component analyses allowed to select a limited number of features in time, frequency and non-linear domains based on independence.

Data were expressed by the median and the 25th and 75th percentiles. We used the Kruskal–Wallis test for inter group comparisons and the Mann–Whitney U–test for intragroup comparisons with a threshold at 0.05 (Siegel and Castellan, 1989). PCA on quantile normalized values from the analyses of the center of pressure and the DHI was used to identify groups of correlated features.

Possible scores on the DHI range from 0 to 100, with higher scores indicating greater self-reported difficulty with dizziness and/or imbalance. The DHI scores for the control group ranged from 0 to 4, (0;0–5.5) while the DHI scores for the PD group ranged from 0 to 44 (6, 0–14) (P < 0.05). Two participants in the control group indicated intermittent difficulty on two separate DHI questions. One control participant indicated intermittent dizziness or imbalance with quick movements of the head (item 11), and the other participant indicated intermittent dizziness or imbalance when bending over (item 25). These results are in stark contrast to the DHI results from the PD group. About two\third of the PD population reported score below the pathological threshold (60%, score < = 6) while the third of participants in the PD group (40%) scored above pathological threshold (score > 6). In the PD population, a majority of patients (60%) reported at least some level of difficulty with dizziness and/or imbalance on multiple items of the DHI. Twenty-one of the 25 DHI items were noted as problematic for at least some of the participants with PD (see Supplemental table 1); however, the difficulties noted by the participants with PD was quite variable across DHI items. The four items most frequently noted by the participants with PD to cause difficulty were items 11, 19, 25, and 5. Items 11 and 25 are within the physical subscale of the DHI, which assesses factors that precipitate and exacerbate symptoms of dizziness and/or imbalance. These are the same two items identified by two participants in the control group (quick head movements and bending over); however, a greater number of participants in the PD group noted difficulty with these situations. Participants with PD also frequently identified limitations with ambulating in the dark (item 19) and with getting into and out of bed (item 5), suggesting functional limitations in the daily activities. Difficulty with items on the emotional subscale of the DHI was noted by a small number of PD participants. These participants reported feeling embarrassed (item 10), frustrated (item 2), handicapped by their dizziness or imbalance (item 21), and afraid others may perceive them as intoxicated (item 15). A few of the participants with PD also reported that their dizziness and/or imbalance caused them to restrict participation in social activities (item 6) and travel for business or recreation (item 3), as well as to experience functional limitations related to household and job responsibilities (items 14 and 24). Overall, there was variability in the items noted by the participants in the PD group as causing difficulty. These results suggest that individuals with PD experience functional limitations and negative emotional responses to dizziness and/or imbalance experienced; however, the exact effects of the dizziness and/or imbalance is not standard across individuals with PD.

All control subjects were able to perform the four mCTSIB conditions for the duration of 60 s. In the PD group, all subjects performed well with normal proprioception (firm surface) and 81.9% of them succeeded the standing test with sensorial deprivation. Only 18.2% of PD patients lost their balance before 60 s when tested on the foam (see Supplemental table 2). For those patients who lost balance on the foam, the loss of visual input drastically reduced their standing duration from 23.33 s (8.9–37.8) to 4.52 s (4.45–4.6) (p < 0.05) (see Supplemental table 3).

Screen testing strategy used in this study was based on a multivariate analysis of subjective measurement (DHI scoring) and objective measurement of features in the time, frequency, and non-linear domain calculated standing balance performance and the dynamic of the center or of pressure collected during the paradigms tested.

The Euclidian space defined by the first three components of the PCA is presented Figure 1 and the abbreviation used to label the binomial (paradigm, feature) are indicated in Table 1. The three first components explained individually more than 15% of the variance and, combined, 63% of the variance (see Table 2).

The contributions (weight) of each variable in the first three principal components are given in Table 3. The first component (C1, 26.24%) opposed irregularity and duration (FCApEn, FOApEn, FCT > 0.25; F:foam, C: closed eye, O: open eyes, ApEn: irregularity, T: duration of standing) to oscillations and dispersion (FCMAD,FCHF,FCLF < −0.25; MAD: median absolute deviation, HF: high frequency, LF: low frequency) mostly for paradigms on with combined sensorial deprivation (foam and closed eyes).

The second component opposed irregularity and high frequency oscillations (HCApEn, FOHF, HOApEn > 0.3) to low frequency and dispersion (HOLF, HCLF, FOMAD < −0.3) mostly on the firm surface.

Subjective measurement of balance function poorly contributed poorly to these two first components as indicated by the weight of the DHI scoring contained between 0.15 and −0.15 for both C1 and C2.

In contrast, component 3 presented relevant clinical significance as it opposed self-reported symptoms (SRS > 0.25) to duration (T < −0.11). C3 also opposed dispersion and high frequency (HCMAD, HCHF, HOHF, FCHF, FOMAD > 0.25) to LF and ApEn (FOLF, HCApEn < −0.15). Paradigms with visual deprivations mostly contributed to this component (HCMAD, HCHF, HOHF, FCHF opposed to HCApEn, FCT).

Analyses of the selectivity and specificity between the three first components to differentiate PD subjects from healthy subjects showed, unsurprisingly, that component 3 discriminated best these two populations (Figure 1C, sensitivity: 0.67, specificity: 0.71; Figure 1).

We have investigated a new paradigm to screen patients with parkinsonism based on a multivariate analyses of subjective and objective measurements of the balance function. The system includes features from the DHI questionnaire and features of on the dynamic of the center of pressure monitored during a 4-paradigm standing test and using a low cost force board. An automated analyses based on principal component allowed to achieve a sensitivity of 0.67 and selectivity of 0.71.

In PD population, the obvious variability in score to the DHI and self-reported dizziness, unsteadiness, or imbalance may relate to many factors ranging from severity of the disease (Lee et al., 2016), age (Harris et al., 2015; Venhovens et al., 2016), response to treatment (Odin et al., 2015), occupational activity (Frazzitta et al., 2015), and awareness on the condition (van der Kooij et al., 2007). Another possible explanation is that variability among patients may exist on the time course of lesion along the circuitries related to balance functions resulting in large variability on subjective balance deficit. These findings support the views that routine screening is needed among the population with risk for parkinsonism.

About half the population of PD patients reported balance deficit in a pathological range (DHI score above six). This result shows the relevance of subjective measurement to screen for parkinsonism (Leddy et al., 2011) and also the need of complementary objective measurements for those unaware of their deficits (van der Kooij et al., 2007; Kahle and Highsmith, 2014). Principal component analyses shows that for the most discriminating component, objective measurements exhibit stronger weigh than the DHI and reinforce previous investigation suggesting that subjective and objective measurement are rather complementary in the appreciation of balance deficits (van der Kooij et al., 2007; Rossi-Izquierdo et al., 2015).

The use of paradigms with sensorial deprivation appeared as an important point of our system as shown by the weight of some features calculated on the center of pressure in conditions with sensorial deprivations. In addition, a small subpopulation of PD participants show reduced functional balance only when proprioceptive input is unreliable, which is compounded when visual input is eliminated. Healthy controls did not show the same difficulty maintaining balance on a compliant surface, with and without visual input. The most discrimant component showed positive relationship between self-reported symptoms, dispersion and high frequency; these features were opposed to duration of standing, irregularity, and slow oscillation. This finding suggests that PD patients and healthy subjects differ in their strategy to compensate sensorial deprivation. The compensatory mechanisms underlying these findings may be major contributors to patient unawareness on their individual balance deficits and point on the importance to “bypass” these compensatory mechanisms when testing balance functions in this population. In addition, this finding shows that vestibular dysfunction is likely a relevant feature to test screen subject with high risk for parkinsonism. In regard to Braak hypothesis on the predominance of peripheral lesions in the early stage of parkinsonism (Braak et al., 1995, 1996, 2002, 2005; Braak and Braak, 2000; Del Tredici et al., 2002; Burke et al., 2008; Fujishiro et al., 2008), the vestibular organ and its connection to central balance circuitry need to be further investigated as a putative hallmark for early diagnosis.

The multivariate analyses on subjective measurement of balance deficit and on objective measurement of center of pressure during a 4 paradigm tests allowed discrimination between PD patients (vs. healthy control) with a selectivity of 0.67 and a specificity of 0.71. Inarguably, these statistical measures show the relevance of this testing strategy for screening purposes but not for diagnosis. The low duration of the testing (about 10 min), the automatization of the analyses (that reduces need of trained staff) and low cost of equipment highlight the cost-effectiveness of such strategy for screen testing patients with parkinsonism. Such screening platform are good putative to meet the need for broad screen testing for PD in both private or public practice and especially in areas with social-economical constraint.

While some studies have reported that game industry boards can adequately estimate the center of pressure (Clark et al., 2010; Holmes et al., 2013; Huurnink et al., 2013; Goble et al., 2014), others have reported some limitations (Wikstrom, 2012). Our protocol may certainly benefit from being interfaced with equipment that are technologically more advanced to monitor balance (Rahmatullah et al., 2015). However, the fact that this material is already broadly used in recreational activity facilitate its implementation in clinical environment and especially those with limited resources. For example, game industry boards have already gain a lot of attention in physical therapy (Goble et al., 2014; Harris et al., 2015) for training purposes; some of these training software's are now available for personal use at home (Esculier et al., 2012). Shortfalls of using balance board from game industry may include some limitations in the selectivity and sensitivity of the screening test and, certainly, some limitation in the utilization of the data collected for research purposes. Alternatives to the use of professional balance board may exist and include the implantation of digital filters to reduce the instability of sensors, the identification of features more robust to the limitation of the equipment's and the use of machine learning to improve the strategy of discrimination.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.