12. Ridgel, 2013

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 21, NO. 3, MAY 2013 481

Variability in Cadence During Forced CyclingPredicts Motor Improvement in Individuals

With Parkinson’s DiseaseAngela L. Ridgel, Hassan Mohammadi Abdar, Student Member, IEEE, Jay L. Alberts, Fred M. Discenzo, and

Kenneth A. Loparo, Fellow, IEEE

Abstract—Variability in severity and progression of Parkinson’sdisease symptoms makes it challenging to design therapy inter-ventions that provide maximal benefit. Previous studies showedthat forced cycling, at greater pedaling rates, results in greaterimprovements in motor function than voluntary cycling. Theprecise mechanism for differences in function following exer-cise is unknown. We examined the complexity of biomechanicaland physiological features of forced and voluntary cycling andcorrelated these features to improvements in motor functionas measured by the Unified Parkinson’s Disease Rating Scale(UPDRS). Heart rate, cadence, and power were analyzed usingentropy signal processing techniques. Pattern variability in heartrate and power were greater in the voluntary group when com-pared to forced group. In contrast, variability in cadence washigher during forced cycling. UPDRS Motor III scores predictedfrom the pattern variability data were highly correlated to mea-sured scores in the forced group. This study shows how time seriesanalysis methods of biomechanical and physiological parametersof exercise can be used to predict improvements in motor function.This knowledge will be important in the development of optimalexercise-based rehabilitation programs for Parkinson’s disease.

Index Terms—Exercise, motor function, movement disorders,neurorehabilitation, signal processing.

NOMENCLATURE

UPDRS Unified Parkinson’s Disease RatingScale.

Stdev Standard deviation.

Approximate entropy.

Spectral entropy.

Manuscript received January 26, 2012; revised June 12, 2012; acceptedSeptember 30, 2012. Date of publication October 24, 2012; date of currentversion May 04, 2013. This work was supported by the National Institutes ofHealth under Grant NIH R21HD068846 (ALR) and Grant NIH R21HD056316(JLA).A. L. Ridgel is with the Department of Exercise Science, Kent State Univer-

sity, Kent, OH 44242 USA.H. M. Abdar and K. A. Loparo are with the Department of Electrical Engi-

neering and Computer Science, Case Western Reserve University, Cleveland,OH 44106 USA.J. L. Alberts is with the Department of Biomedical Engineering, Center for

Neurological Restoration, Cleveland Clinic, Cleveland, OH 44195 USA andalso with the Cleveland FES Center, L. Stokes Cleveland VA Medical Center,Cleveland, OH 44106 USA.F. M. Discenzo is with the Advanced Technology Laboratory, Rockwell Au-

tomation, Mayfield Heights, OH 44124 USA.Digital Object Identifier 10.1109/TNSRE.2012.2225448

Sample entropy.

Probability of a 1 (the proportion of 1 s,the mean of ).

Odds ratio.

Exponent of the odds ratio.

RPM Revolutions per minute.

PSD Power spectral density.

SRM Schoberer Rad Meßtechnik.

BWSTT Body weight supported treadmilltraining.

I. INTRODUCTION

P ARKINSON’S disease (PD), which affects nearly 1.5 mil-lion Americans [1], is a progressive neurological disorder

that is characterized by the loss of dopaminergic neurons in thebrainstem. As PD progresses, the combined motor (e.g., tremor,bradykinesia, rigidity, posture, and gait disorders) and nonmotorsymptoms (e.g., cognitive deficits, depression) often lead to de-creased independence and increased reliance on others for ac-tivities of daily living. Dopaminergic drugs (e.g., levodopa) re-duce tremor, bradykinesia, and rigidity symptoms but do notslow down the progression of the disease. As medication doselevels increases over time, motor fluctuations and dyskinesiaoften develop. Furthermore, these medications are costly andcan cause side effects such as nausea, orthostatic hypotension,and drowsiness [2]. There is a need for additional therapies thatimprove motor function. Exercise and movement therapies havebeen shown to benefit individuals with PD but there is little con-sensus on the optimal mode or intensity [3]–[5].We previously developed a novel approach to increase exer-

cise intensity in individuals with PD called forced exercise [6].This approach used a stationary tandem bicycle and an able-bodied cyclist to assist individuals with PD to pedal at a ca-dence [revolutions per minute (rpm)] between 80–90 rpm. Thiscadence was roughly 30% faster than they were able to pedalon their own at a self-selected rate that was determined duringbaseline evaluation. Crank-based powermeters (Schoberer RadMeßtechnik [SRM]) measured the work of the individuals withPD during each exercise session. Forced exercise results in asignificant improvement in motor symptoms as measured withthe Unified Parkinson’s Disease Rating Scale (UPDRS) Motor

1534-4320/$31.00 © 2012 IEEE

482 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 21, NO. 3, MAY 2013

III. This clinical scale evaluates the degree of tremor, brady-kinesia, rigidity, and posture/gait difficulties in individuals withPD. However, individuals that cycled on a single stationary er-gometer at similar aerobic intensity, but at a self-selected ped-aling rate (voluntary exercise) showed no change in UPDRSMotor III scores.The goal of this study was to examine the complexity of

biomechanical and physiological features of forced and vol-untary cycling and to relate these features to improvements inmotor function as measured by the UPDRS Motor III scale. Wehypothesize that temporal variability or lack of predictability incadence during forced cycling can be used to accurately predictresulting improvements in UPDRS Motor III scores.Subject and trainer data (power, heart rate, and cadence), col-

lected previously from exercise training sessions using a sta-tionary tandem bicycle in our previous study [6], were examinedusing sophisticated signal processing techniques: approximateentropy , sample entropy , and spectral en-tropy . is a “regularity statistic” that quanti-fies the unpredictability of temporal fluctuations in a time se-ries such as an instantaneous heart rate time series, . Thepresence of repetitive temporal patterns in a time series rendersit more predictable than a time series in which such patterns areabsent. quantifies the likelihood that “similar” patternsof observations will not be followed by additional “similar” ob-servations. A time series containingmany repetitive patterns hasa relatively small ; a less predictable (i.e., more randomor less time-correlated) time series will have a greater .

is a modification of [7] that removes the po-tential bias in [8], eliminates self-matches in the compu-tation, reduces computational complexity, and can be applied toshort time series data. Both and quantify thepredictability (or regularity) in a time series, and are useful inquantifying differences in health and disease [7]–[11]; whereas

quantifies the distribution of frequency content in atime series.More sophisticated analysis of the exercise performance vari-

ables across groups and throughout exercise sessions will aid inidentifying and quantifying time and frequency domain featuresthat may be responsible for the improved motor performanceobserved after forced exercise. Our previous studies showed thatbehavioral effects of forced and voluntary exercise were dra-matically different [6]. A precise understanding of specific anddifferentiating characteristics between forced and voluntary ex-ercise will provide important guidance in the development ofmore cost and time effective methods of delivering forced ex-ercise than tandem cycling, such as motorized single bikes thatwill not require a trainer [12].

II. MATERIALS AND METHODS

A. Forced and Voluntary Cycling Data CollectionTen individuals with idiopathic Parkinson’s disease were

assigned to one of two groups: forced (tandem) or voluntary(single) cycling. Both groups completed 24 1-h exercise ses-sions (three per week) over an eight-week period. In the forcedcycling group, the PD subject was assisted by an able-bodiedcyclist and Certified Personal Trainer riding on the front of

a stationary tandem bike. The trainer had the objective ofmaintaining bike operation at an accelerated cadence ratebetween 80 and 90 rpm. The voluntary cycling group pedaled astationary single bike (SRM indoor trainer, Jülich, Germany) ata self-selected cadence of roughly 60 rpm. The pedal cadenceand power performed by the subject and the trainer on thetandem and on the single bicycle were measured using SRMpowermeters. A Polar heart rate monitor (Polar Electro, LakeSuccess, NY, USA) was used to collect heart rate data. Heartrate (HR), cadence, and power variables were averaged for20–24 sessions per subject. In order to examine the raw datafor single (voluntary) and tandem (forced) rider exercise tests,the mean and standard deviation of the “heart rate,” “power,”and “cadence” signals for all tests were calculated. Approx-imately 20 or more data sets for each person were collectedfor exercise sessions across the eight-week intervention. Addi-tional methodological details and photos of the tandem setupcan be found in Ridgel et al. 2009 [6]. The Cleveland ClinicInstitutional Review Board approved this project.

B. Motor Function AssessmentThe Unified Parkinson’s Disease Rating Scale (UPDRS) Part

III motor exam was administered while individuals were “off”anti-Parkinsonian medication for 12 h. Assessments were per-formed prior to and after completion of the eight week exer-cise intervention. The difference in UPDRS Motor III scoresbetween these two time periods was calculated and was usedin the model (described below). A positive score represents im-provement while a negative score indicates worsening of motorsymptoms.

C. Biomechanical and Physiological Feature AnalysisThree main parameters approximate entropy ,

sample entropy , and spectral entropyfor “heart rate,” “power,” and “cadence” signals were com-puted in all data sets for each person. These parameters wereused to quantify complexity in terms of both temporal andfrequency domain patterns of variability. The values forand computed for the dataset are approximatelythe same, so further analysis focused on the computa-tions. The mean of these parameters for all exercise sessionswere computed for the single (voluntary) and tandem (forced)groups.Approximate entropy was used to quantify the de-

gree of temporal regularity or predictability in the time seriesdata for power, heart rate, and cadence. The algorithm for com-puting has been published elsewhere [8], [9]. Here, weprovide a brief summary of the calculations, as applied to atime series of instantaneous heart rate measurements, .Given a sequence , consisting of instantaneous heart ratemeasurements , and a time delay

, we chose values for two input parameters, and ,to compute the approximate entropy, , of thetime series. The parameter specifies the pattern length, andthe parameter defines the criterion of similarity. We denoteda subsequence (or pattern) of heart rate measurements, be-ginning at measurement within , by the vector . Twopatterns, and , are similar if the difference between

RIDGEL et al.: VARIABILITY IN CADENCE DURING FORCED CYCLING PREDICTS MOTOR IMPROVEMENT IN INDIVIDUALS WITH PD 483

TABLE IMEAN AND VARIANCE FOR POWER, HEART RATE, AND CADENCE SIGNALS

positive change in UPDRS represents improvements in motor functionmean values were calculated over 20–24 exercise sessions per patient

any pair of corresponding measurements in the patterns is lessthan the tolerance , i.e., if

Now consider the set of all patterns of length [i.e.,], within . We now

defined

where is the number of patterns in that are similarto (given the similarity criterion ). The quantityis the fraction of patterns of length that resemble the patternof the same length that begins at interval . We can calculate

for each pattern in , and we define as themean of these values. The quantity expresses theprevalence of repetitive patterns of length in . Finally, wedefine the of , for patterns of length and similaritycriterion , as

Note, is the natural logarithm of the rela-tive prevalence of repetitive patterns of length compared withthose of length .

estimates the logarithmic likelihood that the next inter-vals after each of the patterns will differ (i.e., that the similarityof the patterns is mere coincidence and lacks predictive value).Smaller values of imply a greater likelihood that similarpatterns of measurements will be followed by additional similar

TABLE IIAPEN AND SPECEN FOR POWER, HEART RATE, AND CADENCE SIGNALS

measurements. If the time series is highly irregular, the occur-rence of similar patterns will not be predictive for the followingmeasurements, and will be relatively large. It should benoted that has significant weaknesses, notably its strong


TABLE IIIREGRESSION ANALYSIS RESULTS FOR SINGLE AND TANDEM RIDER TESTS

dependence on sequence length and its poor self-consistency(i.e., the observation that for one data set is larger than

for another for a given choice of and , does not nec-essarily, hold true for other choices of and ).Sample entropy was used to quantify the com-

plexity of a time series using a measure of predictability thatis very similar to . Spectral entropy was usedto measure the complexity of time series data in the frequencydomain and was used to quantify frequency domain variabilityin power by computing the power-spectral-density (PSD) ofthe time series. The PSD was normalized to produce a proba-bility-like density function and transformed with the Shannonfunction as follows.1) Compute the power spectral density (PSD) of the signal,

.2) Normalize the power spectrum

3) Compute the Shannon function

4) Compute the spectral entropy

where is the number of data points.Statistical analysis, including linear regression, logistic

regression and computation of the odds ratio, was used toinvestigate the relationship between the measures of exercise

variability and the exercise related change in motor perfor-mance as measured by UPDRS Motor III scores. The data inTables I and II were used to develop a multiple linear regressionmodel (MLR) and apply logistic regression to determine theodds ratio for achieving a positive change in the UPDRS MotorIII scores in each group. Each regression model has four inde-pendent variables: (heart-rate), (power),(cadence), and (power), and one dependent variable:UPDRS score. The data was analyzed in two ways: first twoseparate MLR models were built, one model for the singlesessions and one model for the tandem sessions (Table III)and then the data were combined into a single dataset and asingle MLR model was built (Table IV). The residual values inTables III and IV, and the predicted UPDRS values in Fig. 1were obtained fromMLRmodeling using the “regress” functionin MATLAB. The odds ratio is the ratio of the odds of an eventoccurring in one group to the odds of it occurring in anothergroup. It is also used to refer to sample-based estimates of thisratio. The predicted UPDRS Motor III scores were obtainedusing MLR as described above. A Pearson product-momentcorrelation coefficient was computed to assess the relationshipbetween the real and predicted values of UPDRS motor IIIscores.Dependent variables (e.g., heart rate, power, cadence, power

, heart rate , cadence , power ) werecompared between single and tandem groups using a one-wayANOVA (SPSS, Inc., ver. 18). Significance was set at .

III. RESULTS

There were no significant differences in heart rate andpower between the two groups (Table I). However, the ped-aling cadence showed a significant difference ( ,


TABLE IVCOMBINED SINGLE AND TANDEM RIDER DATA REGRESSION ANALYSIS RESULTS

) between the raw values in voluntary (60.8 12.3rpm) and forced (84.5 2.4 rpm) groups. Cadence for theforced exercise sessions was higher than the voluntary sessionswith less variability as quantified by the standard deviation.For “ ,” only the power signals are distinguishable be-

tween the single and tandem groups. Comparison of each vari-able shows clear differences. for the power in the singlesessions (0.247 0.10) is significantly greater ( ,

) than the for the power in the tandem sessions(0.032 0.02). for the heart rate in single sessions (0.3900.09) is also significantly greater ( , )

than the for the heart rate in the tandem group (0.1450.02). This suggests that the power and heart rate signals in

the voluntary (single) group have greater variability (are lesspredictable) than the signals in forced (tandem) group. The re-sults for the cadence signal are opposite; that is, the cadencesignals for the single group (0.378 0.18) show less variability(are more predictable) and are significantly less ( ,

) than the cadence signals for tandem sessions (1.010.15). Spectral entropy of the power in the tandem

group (0.139 0.05) showed slightly greater, but not signifi-cant, variability ( , ) than in the singlesessions (0.089 0.01).Predicted values of UDPRS, using MLR analysis, are plotted

against real (measured) values in Fig. 1. Four out of the five in-dividuals who completed single (voluntary) sessions showed noimprovement or worsening of UPDRS Motor III scores whileone individual showed a slight improvement [Fig. 1(a)]. TheMLR model is less accurate in individuals whose scores wors-ened. All participants who completed forced exercise sessionsshowed improvement in the UPDRSMotor III scores [Fig. 1(b)]and the model shows a more accurate prediction in this group.

The combined model results in greater differences between thepredicted and real scores for most subjects [Fig. 1(c)].There was no significant correlation between the real and

predicted UPDRS Motor III scores in the voluntary exercise(single) group (Fig. 2(a), , , ). How-ever, there was positive and significant correlation in the forcedexercise (tandem) group (Fig. 2(b), , ,

). When the two groups were combined, there was also asignificant and positive correlation between real and predictedUPDRS values (Fig. 2(c), , , ).

IV. DISCUSSION

Although many studies have documented the benefits of ex-ercise, it is unclear what elements (i.e., dosage, intensity, inter-vention type) constitute an optimal exercise intervention for PD.Each individual with PD has different symptoms and capabili-ties that make it challenging to design a single rehabilitationprogram that is optimal for all. Furthermore, progression of thedisease often requires reassessments and changes to rehabilita-tion programs.This study reveals that pattern irregularity in HR and power

is greater in the single sessions when compared to the tandemsessions, indicating that the trainer provides a “stabilizing” in-fluence on the patient’s exercise intensity, while maintaining el-evated cadence. The single PD rider has a tendency to intro-duce greater variability (less regular patterns) in power output,inducing greater fluctuations in HR, when compared to tandemPD riders. In contrast, the cadence signal shows greater vari-ability during the tandem sessions. This variability is likely dueto the inability of individuals with PD to maintain a constanthigh-speed pedal cadence. Furthermore, variability was also in-troduced when the able-bodied trainer was required to increase


Fig. 1. Real (measured) versus predicted UPDRS values. (A) Single (volun-tary) exercise data. (B) Tandem (forced) data. (C) Combined single and tandemdata. A multiple linear regression model was used to calculate predicted valuesof UPDRS. The model more accurately predicted UPDRS Motor III scores inthe tandem (forced) group than in the single (voluntary) group. The combinedmodel was inaccurate in predicting scores in both groups.

or decrease pedal speed to maintain the desired cadence. Thesingle PD riders rode at a self-selected cadence and thus showedlower variability during exercise bouts.This supports our hypothesis that temporal variability or lack

of predictability (not quantified by conventional statistical pa-rameters such as variance or coefficient of variation) in cadence

Fig. 2. Correlation analysis of real (measured) versus predictedUPDRS values.(A) Single (voluntary) exercise data. (B) Tandem (forced) data. (C) Combinedsingle and tandem data. There was a positive and significant correlation of thereal and predicted UPDRS Motor III scores in the tandem group but no correla-tion in the single group. Combined scores maintained a significant correlationbetween real and predicted scores.

during forced exercise can be used to accurately predict re-sulting improvements in UPDRS Motor III scores. Lastly, pre-dicted UPDRS Motor III scores are highly correlated to mea-sured scores in the tandem sessions. These data provide insightinto how times series analysis methods can be applied to un-cover potential features in the measured variables and how this


information can be used to correlate exercise parameters withimproved motor function.

A. Changes in the Complexity of Motor Output With PD

Typically, researchers use the mean and standard deviationor standard error of the mean to define variability in a dataset.These measures provide a description of the magnitude of thevariability around a central point. However, the presence of cer-tain patterns or shifts in patterns can often provide primary in-sight into health status or motor performance [13], [14]. Assess-ment of irregularities of serial data using entropy statistics hasbeen shown to reveal subtle disruptions in movement patternsprior to changes in mean and variance.Previous work has postulated that aging and disease are

accompanied by reduction in the complexity of physiologicaland behavioral control [15], [16]. This loss of complexity canreduced the body’s ability to adapt to physiological stress.For example, Vaillancourt et al. [17] examined the ofhand tremor during a grip force task in individuals with PDand healthy age-matched controls. They showed that tremor isless variable in PD than healthy controls and that there was anegative correlation between variability of tremor and severityof the disease, as measured by the UPDRS Motor III. Thissuggests that progression of PD results in decreased variabilityof motor output. In light of these findings, it is possible thatexercise or movement training, that emphasizes complex andvariable movements, could promote motor improvement in PD.

B. Role of Intensity and Complexity of Exercise inImprovements

Few other studies have correlated the biomechanics of exer-cise with improved function in PD. Abe et al. examined rota-tional velocities and relative phase during leg pedaling in in-dividuals with PD and were able to divide their subjects intogroups using cluster analysis [18], [19]. Each group showed dis-tinctive clinical features, specifically the probability of freezingof gait.Several studies have described how activity-dependent neu-

roplasticity of the motor cortex and basal ganglia could be af-fected by the intensity and complexity of motor training andexercise [20]–[27]. Body weight supported treadmill training(BWSTT) has been used to increase intensity of exercise by al-lowing individuals to walk faster than they are able over-ground.This paradigm uses a harness to support a small percentage ofthe body weight. After 24 sessions of BWSTT, subjects showedimproved walking performance and UPDRS Motor III scores[24], [27], [28]. Furthermore, Fisher et al. documented length-ening of the cortical silent period using transcranial magneticstimulation (TMS) after BWSTT but not after standard phys-ical therapy [21]. The authors suggested that improvements afterBWSTT was due to the fact that it allowed higher treadmillspeed compared to walking without body weight support. Thesefindings support the idea of dose-dependent benefits of exerciseand suggest that high-intensity exercise can normalize cortico-motor excitability in early PD.In animal models of PD, high intensity treadmill running re-

sulted in gait and balance improvements, as well as partial re-

covery of dopaminergic neurons after injection of the neuro-toxin, MPTP [20], [26], [29], [30]. Animals that exercised ona running wheel at their own pace or those that did not exer-cise did not show similar improvements. In addition, Petzingersuggested that intensive treadmill running resulted in differentneuroplastic changes within the basal ganglia between lesionedand nonlesioned animals [29]. Despite these promising studies,there is still much to be learned about the specific exercise pa-rameters that can maximize motor improvement in PD.

C. Implications for Neurorehabilitation

Although high-intensity exercise can improve motor functionin PD and animal models of PD, the mechanisms of this im-provement are unclear. Understanding of potential mechanismswill be important in the development of additional rehabilitationparadigms that could benefit these individuals. It is possible thatchanges in proprioceptive input during a bout of tandem cyclingprovide complex and variable sensory feedback that increasescortical activation [3].Accurate voluntary movements require somatosensory input

from the periphery. Peripheral receptors, such as joint recep-tors, golgi tendon organs, muscle spindles, and cutaneous re-ceptors, send feedback information from the limbs to the motorcortex. Several studies have identified proprioceptive impair-ment in PD, specifically in muscle spindle responses, load sen-sitivity, and kinesthesia [31]–[34]. However, levodopa does notappear to improve kinesthetic deficits in PD [35], [36] and hasbeen associated with suppression of sensitivity to joint position[37]. Therefore, deficits in peripheral afferent input or in the cor-tical response to that input could affect motor output in individ-uals with PD.During leg cycling, there are a number of proprioceptive sig-

nals including hip and knee joint angle, muscle length and force,as well as cutaneous input from the bottom of the foot [38]. Im-provements in motor function and mobility after bouts of cy-cling in individuals with PD could be due to increases in afferentinput to the cortex. Snijders et al. described a case in whichan individual with PD presented with severe freezing of gaitbut was able to ride a bicycle [39], [40]. They suggested thatforces transferred from the rotating pedals to the individual’sfeet might provide tactile cues that trigger appropriate rhythmicmovements to overcome freezing of gait. This suggests that acti-vation of proprioceptors with a high frequency but variable pat-tern may be important for motor improvements in PD. Futurestudies should assess the role of proprioceptive input in motorimprovements after forced cycling by altering or removing thatinput.

ACKNOWLEDGMENT

A. L. Ridgel jointly conceived and wrote this study withK. A. Loparo and F. M. Discenzo. H. M. Abdar completedthe results for the variability data analysis. The original forcedexercise study was executed and written by A. L. Ridgel.J. L. Alberts provided equipment, space, and guidance forthe forced exercise study and provided the raw data for thisanalysis. All authors reviewed this manuscript.


REFERENCES

[1] E. R. Dorsey, R. Constantinescu, J. P. Thompson, K. M. Biglan, R.G. Holloway, K. Kieburtz, F. J. Marshall, B. M. Ravina, G. Schifitto,A. Siderowf, and C. M. Tanner, “Projected number of people withParkinson disease in the most populous nations, 2005 through 2030,”Neurology, vol. 68, no. 5, pp. 384–386, Jan. 30, 2007.

[2] O. Rascol, P. Payoux, F. Ory, J. J. Ferreira, C. Brefel-Courbon, and J.L. Montastruc, “Limitations of current Parkinson’s disease therapy,”Ann. Neurol., vol. 53, no. Suppl 3, p. S3-12, 2003, discussion S12-5.

[3] J. L. Alberts, S. M. Linder, A. L. Penko, M. J. Lowe, and M. Phillips,“It is not about the bike, it is about the pedaling: Forced exerciseand Parkinson’s disease,” Exer. Sports Sci. Rev., vol. 39, no. 4, pp.177–186, Oct. 2011.

[4] T. Archer, A. Fredriksson, and B. Johansson, “Exercise alleviatesParkinsonism: Clinical and laboratory evidence,” Acta NeurologicaScand., vol. 123, no. 2, pp. 73–84, 2011.

[5] S. H. Keus,M.Munneke,M. J. Nijkrake, G. Kwakkel, and B. R. Bloem,“Physical therapy in Parkinson’s disease: Evolution and future chal-lenges,” Mov. Disord., vol. 24, no. 1, pp. 1–14, Jan. 15, 2009.

[6] A. L. Ridgel, J. L. Vitek, and J. L. Alberts, “Forced, not voluntary, ex-ercise improves motor function in Parkinson’s disease patients,” Neu-rorehabil. Neural Repair, vol. 23, no. 6, pp. 600–608, Jul.–Aug. 2009.

[7] J. S. Richman and J. R. Moorman, “Physiological time-series analysisusing approximate entropy and sample entropy,” Am. J. Physiol. HeartCirc. Physiol., vol. 278, no. 6, pp. H2039–H2049, Jun. 2000.

[8] S. M. Pincus, “Approximate entropy as a measure of system com-plexity,” Proc. Nat. Acad. Sci. USA, vol. 88, no. 6, pp. 2297–2301,Mar. 15, 1991.

[9] F. Kaffashi, R. Foglyano, C. G. Wilson, and K. A. Loparo, “The effectof time delay on approximate sample entropy calculations,” Physica D,vol. 237, pp. 3069–3074, 2008.

[10] J. S. Richman, D. E. Lake, and J. R. Moorman, “Sample entropy,”Methods Enzymol., vol. 384, pp. 172–184, 2004.

[11] S. M. Pincus and R. R. Viscarello, “Approximate entropy: A regularitymeasure for fetal heart rate analysis,” Obstetr. Gynecol., vol. 79, no. 2,pp. 249–255, Feb. 1992.

[12] A. L. Ridgel, C. A. Peacock, E. J. Fickes, and K. C.-H. , “Active-as-sisted cycling improves tremor and bradykinesia in Parkinson’s dis-ease,” Arch. Phys. Med. Rehabil..

[13] S. M. Pincus, “Assessing serial irregularity and its implications forhealth,” Ann. NY Acad. Sci., vol. 954, pp. 245–267, Dec. 2001.

[14] N. Stergiou and L. M. Decker, “Human movement variability, non-linear dynamics, and pathology: Is there a connection?,” HumanMovem.Sci., vol. 30, no. 5, pp. 869–888, Oct. 2011.

[15] L. A. Lipsitz and A. L. Goldberger, “Loss of ‘complexity’ and aging.Potential applications of fractals and chaos theory to senescence,”JAMA, vol. 267, no. 13, pp. 1806–1809, Apr. 1, 1992.

[16] D. E. Vaillancourt and K. M. Newell, “Changing complexity in humanbehavior and physiology through aging and disease,”Neurobiol. Aging,vol. 23, no. 1, pp. 1–11, Jan.–Feb. 2002.

[17] D. E. Vaillancourt, A. B. Slifkin, and K. M. Newell, “Regularity offorce tremor in Parkinson’s disease,” Clin. Neurophysiol., vol. 112, no.9, pp. 1594–1603, Sep. 2001.

[18] K. Abe, Y. Asai, Y. Matsuo, T. Nomura, S. Sato, S. Inoue, I. Mizukura,and S. Sakoda, “Classifying lower limb dynamics in Parkinson’s dis-ease,” Brain Res. Bull., vol. 61, no. 2, pp. 219–126, Jul. 15, 2003.

[19] Y. Asai, T. Nomura, S. Sato, A. Tamaki, Y. Matsuo, I. Mizukura, andK. Abe, “A coupled oscillator model of disordered interlimb coordina-tion in patients with Parkinson’s disease,” Biol. Cybern., vol. 88, no. 2,pp. 152–162, Feb. 2003.

[20] B. E. Fisher, G. M. Petzinger, K. Nixon, E. Hogg, S. Bremmer, C. K.Meshul, and M. W. Jakowec, “Exercise-induced behavioral recoveryand neuroplasticity in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyri-dine- lesioned mouse basal ganglia,” J. Neurosci. Res., vol. 77, no. 3,pp. 378–390, Aug. 2004.

[21] B. E. Fisher, A. D. Wu, G. J. Salem, J. Song, C. H. Lin, J. Yip, S.Cen, J. Gordon, M. Jakowec, and G. Petzinger, “The effect of exercisetraining in improving motor performance and corticomotor excitabilityin people with early Parkinson’s disease,” Arch. Phys. Med. Rehabil.,vol. 89, no. 7, pp. 1221–1229, Jul. 2008.

[22] A. M. Graybiel, “The basal ganglia: Learning new tricks and loving it,”Curr. Opin. Neurobiol., vol. 15, no. 6, pp. 638–644, Dec. 2005.

[23] M. A. Hirsch and B. G. Farley, “Exercise and neuroplasticity in personsliving with Parkinson’s disease,” Eur. J. Phys. Rehabil. Med., vol. 45,no. 2, pp. 215–229, Jun. 2009.

[24] G. M. Petzinger, B. E. Fisher, J. E. Van Leeuwen, M. Vukovic, G.Akopian, C. K. Meshul, D. P. Holschneider, A. Nacca, J. P. Walsh,and M. W. Jakowec, “Enhancing neuroplasticity in the basal ganglia:The role of exercise in Parkinson’s disease,”Mov. Disord., vol. 25, pp.S141–S145, 2010.

[25] J. L. Tillerson, A. D. Cohen, J. Philhower, G. W. Miller, M. J. Zig-mond, and T. Schallert, “Forced limb-use effects on the behavioral andneurochemical effects of 6-hydroxydopamine,” J. Neurosci., vol. 21,no. 12, pp. 4427–4435, Jun. 15, 2001.

[26] B. A. Smith, N. R. Goldberg, and C. K. Meshul, “Effects of treadmillexercise on behavioral recovery and neural changes in the substantianigra and striatum of the 1-methyl-4-phenyl-1,2,3,6-tetrahydropy-ridine-lesioned mouse,” Brain Res., vol. 1386, pp. 70–80, Apr. 22,2011.

[27] I. Miyai, Y. Fujimoto, H. Yamamoto, Y. Ueda, T. Saito, S. Nozaki,and J. Kang, “Long-term effect of body weight-supported treadmilltraining in Parkinson’s disease: A randomized controlled trial,” Arch.Phys. Med. Rehabil., vol. 83, no. 10, pp. 1370–1373, Oct. 2002.

[28] I. Miyai, Y. Fujimoto, Y. Ueda, H. Yamamoto, S. Nozaki, T. Saito,and J. Kang, “Treadmill training with body weight support: Its effecton Parkinson’s disease,” Arch. Phys. Med. Rehabil., vol. 81, no. 7, pp.849–852, Jul. 2000.

[29] G. M. Petzinger, J. P. Walsh, G. Akopian, E. Hogg, A. Abernathy, P.Arevalo, P. Turnquist, M. Vuckovic, B. E. Fisher, D. M. Togasaki, andM. W. Jakowec, “Effects of treadmill exercise on dopaminergic trans-mission in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-lesionedmouse model of basal ganglia injury,” J. Neurosci., vol. 27, no. 20, pp.5291–5300, May 16, 2007.

[30] J. L. Tillerson, W. M. Caudle, M. E. Reveron, and G.W. Miller, “Exer-cise induces behavioral recovery and attenuates neurochemical deficitsin rodent models of Parkinson’s disease,” Neuroscience, vol. 119, no.3, pp. 899–911, 2003.

[31] S. V.Adamovich,M. B. Berkinblit,W. Hening, J. Sage, andH. Poizner,“The interaction of visual and proprioceptive inputs in pointing to ac-tual and remembered targets in Parkinson’s disease,” Neuroscience,vol. 104, no. 4, pp. 1027–1041, 2001.

[32] M. Maschke, C. M. Gomez, P. J. Tuite, and J. Konczak, “Dysfunc-tion of the basal ganglia, but not the cerebellum, impairs kinaesthesia,”Brain, vol. 126, pt. 10, pp. 2312–2322, Oct. 2003.

[33] J. Konczak, D. M. Corcos, F. Horak, H. Poizner, M. Shapiro, P. Tuite,J. Volkmann, and M. Maschke, “Proprioception and motor control inParkinson’s disease,” J. Motor Behav., vol. 41, no. 6, pp. 543–552,Nov. 2009.

[34] S. Zia, F. W. Cody, and D. J. O’Boyle, “Identification of unilateralelbow-joint position is impaired by Parkinson’s disease,” Clin. Anat.,vol. 15, no. 1, pp. 23–31, Jan. 2002.

[35] E. E. Jobst, M. E. Melnick, N. N. Byl, G. A. Dowling, and M. J.Aminoff, “Sensory perception in Parkinson disease,” Arch. Neurol.,vol. 54, no. 4, pp. 450–454, Apr. 1997.

[36] W. G. Wright, V. S. Gurfinkel, L. A. King, J. G. Nutt, P. J. Cordo,and F. B. Horak, “Axial kinesthesia is impaired in Parkinson’s disease:Effects of levodopa,” Exp. Neurol., vol. 225, no. 1, pp. 202–209, Sep.2010.

[37] P. O’Suilleabhain, J. Bullard, and R. B. Dewey, “Proprioception inParkinson’s disease is acutely depressed by dopaminergic medica-tions,” J. Neurol. Neurosurg. Psychiatry, vol. 71, no. 5, pp. 607–610,Nov. 2001.

[38] M. Ericson, “On the biomechanics of cycling. A study of joint andmuscle load during exercise on the bicycle ergometer,” Scand. J. Re-habil. Med., vol. 16, pp. 1–43, 1986.

[39] A. H. Snijders and B. R. Bloem, “Images in clinical medicine. Cyclingfor freezing of gait,” New Eng. J. Med., vol. 362, no. 13, p. e46, Apr.1, 2010.

[40] A. H. Snijders, I. Leunissen, M. Bakker, S. Overeem, R. C. Helmich,B. R. Bloem, and I. Toni, “Gait-related cerebral alterations in patientswith Parkinson’s disease with freezing of gait,” Brain, vol. 134, pt. 1,pp. 59–72, Jan. 2011.


Angela L. Ridgel received the Ph.D. degree inbiomedical sciences from Marshall University,Huntington, WV, USA. She completed her post-doctoral training in neurobiology at Case WesternReserve University, Cleveland, OH, USA, and inbiomedical engineering at the Cleveland Clinic,Cleveland, OH, USA.She is an Assistant Professor in Exercise Science

at Kent State University, Kent, OH, USA. For the lastnine years, she has been interested in how aging andneurological disorders limits movement in humans.

Her laboratory examines the effects of exercise and motor rehabilitation ther-apies on motor function, cognition, and balance in Parkinson’s disease. She iscurrently funded through a NIH R21 grant.

Hassan Mohammadi Abdar (S’12) received theB.S. degree in electronic engineering and the M.S.degree in control engineering from K. N. ToosiUniversity of Technology, Tehran, Iran. He is cur-rently working toward the Ph.D. degree in systemsand control engineering at Case Western ReserveUniversity, Cleveland, OH, USA.He has worked in industry for eight years in the

area of embedded systems, industrial automation,and signal processing. He has been working onseveral research projects in the field of microgrid

systems, grid-tie inverter, signal processing, and serial data communication atCase Western Reserve University.

Jay L. Alberts is the Director of the ClevelandClinic Concussion Center and holder of the “EdwardF. and Barbara A. Bell Family Endowed Chair.”He is Associate Staff in the Department of Biomed-ical Engineering and the Center for NeurologicalRestoration. He also holds appointments at theFunctional Electrical Stimulation Center at the LouisStokes VA Medical Center and in the Departmentof Biomedical Engineering at Case Western ReserveUniversity. His research is focused on the role of thebasal ganglia in movement control and the role of

surgical and behavioral interventions on cognitive and motor performance.

Fred M. Discenzo received two B.S. degrees inmathematics, the M.S. degree in polymer physics,and the Ph.D. degree in systems and control en-gineering from Case Western Reserve University,Cleveland, OH, USA.He is the Manager of Rockwell Automation’s

Advanced Technology Laboratory, Cleveland, OH,USA, and is a Rockwell Automation Fellow. Hecurrently holds over 60 U.S. patents and has pub-lished many papers spanning artificial intelligence,machinery diagnostics, sensors, control, power

scavenging, and artificial neural networks.Dr. Discenzo represents Rockwell Automation on various university/in-

dustry advisory committees, he serves on multiple conference committees, andis a board member of the Machinery Failure Prevention Technology (MFPT)Society.

Kenneth A. Loparo (F’99) is the Nord Professor ofEngineering in the Case School of Engineering, CaseWestern Reserve University, Cleveland, OH, USA.He has academic appointments in the Departments ofBiomedical Engineering, Electrical Engineering andComputer Science, and Mechanical and AerospaceEngineering. His research interests include stabilityand control of nonlinear and stochastic systems; non-linear filtering with applications to monitoring, faultdetection, diagnosis and reconfigurable control; in-formation theory aspects of stochastic and quantized

systems with applications to adaptive and dual control; the design of digitalcontrol systems; and advanced signal processing techniques for monitoring andtracking of physiological behavior.

Documents

12. Ridgel, 2013