- Email: [email protected]
Opening velocity, a novel parameter, for finger tapping test in patients with Parkinson's disease
Parkinsonism and Related Disorders 15 (2009) 440–444
Contents lists available at ScienceDirect
Parkinsonism and Related Disorders journal homepage: www.elsevier.com/locate/parkreldis
Opening velocity, a novel parameter, for finger tapping test in patients with Parkinson’s diseaseq M. Yokoe a, e, R. Okuno d, T. Hamasaki b, e, Y. Kurachi c, e, K. Akazawa f, S. Sakoda a, e, * a
Department of Neurology, Osaka University Graduate School of Medicine, Osaka, Japan Department of Medical Statistics, Osaka University Graduate School of Medicine, Osaka, Japan c Department of Pharmacology II, Osaka University Graduate School of Medicine, Osaka, Japan d Department of Electrical and Electronics Engineering, Faculty of Engineering, Setsunan University, Osaka, Japan e The Center for Advanced Medical Engineering and Informatics, Osaka University, Osaka, Japan f Department of Biomedical Engineering, Osaka Institute of Technology, Osaka, Japan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 24 June 2008 Received in revised form 19 August 2008 Accepted 1 November 2008
Objectives: A new system consisting of an accelerometer and touch sensor was developed to find objective parameters for the finger tapping (FT) test in patients with Parkinson’s disease (PD). Methods: We recruited sixteen patients with PD and thirty-two age-matched healthy volunteers (HVs). By using this new system, various parameters related to velocity, amplitude, rhythm and number in the FT test were measured in patients with PD and examined in comparison with those of HVs on the basis of the Unified Parkinson’s Disease Rating Scale (UPDRS) FT score. Results: The new system allowed us to measure fourteen parameters of FT movement very easily, and a radar chart showed obvious differences in most of these parameters between HVs and patients with PD. Principal component analysis showed that fourteen parameters were classified into three components: (1) both mean and standard deviation (SD) of both amplitude and velocity, (2) number of FT for 60 s and mean FT interval, and (3) SD of FT interval. The first (velocity- and amplitude-related parameters) and third (rhythm-related parameters) components contributed to discrimination of PD from HVs. Maximum opening velocity (MoV) was the best of these parameters because of its sensitivity and association with the UPDRS FT score. Conclusions: A novel system for the FT test, which is compact, simple and efficient, has been developed. Velocity- and amplitude-related parameters were indicated to be valuable for evaluation of the FT test in patients with PD. In particular, we first propose that MoV is a novel marker for the FT test. Ó 2008 Elsevier Ltd. All rights reserved.
Keywords: Parkinson’s disease Finger tapping Accelerometer Principal component analysis
1. Introduction Although four clinical symptoms are well known in Parkinson’s disease (PD) – tremor, rigidity, bradykinesia/akinesia and absence of postural reflexes – the nature of each symptom [1] and how symptoms such as rigidity affect movement in patients with PD are still unclear. From a scientific viewpoint to understand the nature of PD movement, it is important to analyze movement from the aspects of rhythm, velocity and amplitude. From a clinical viewpoint, an objective system for assessment of clinical symptoms is required because quantification and consistency in the assessment of PD severity would be of considerable help in clarifying the q The review of this paper was entirely handled by an Associate Editor, Prof. Eng-King Tan. * Corresponding author at: Department of Neurology, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka, Japan. Tel.: þ81 6 6879 3571; fax: þ81 6 6879 3579. E-mail address: [email protected] (S. Sakoda). 1353-8020/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.parkreldis.2008.11.003
efficacy of various drugs. We have focused on the UPDRS finger tapping (FT) score and reported a prototype method for its objective and quantitative evaluation [2,3] because FT as a parameter of upper limb motor function [4] has been widely used to assess motor impairment in patients with neurological or neuropsychological disorders. Accurate scoring and interpretation of results require experience, and even experienced neurologists are unable to detect subtle changes. Goetz and colleagues reported that the UPDRS FT score was one of the most difficult items to assess [5]. Accuracy and reproducibility in the quantitative analysis of FT performance are indispensable for evaluating the severity of PD. The FT test has been analyzed by various methods including the use of a computer keyboard [6–8], a metal plate [9–11], a telegraph key [12] or a triaxial accelerometer [13] for frequency and/or rhythm of FT, a musical instrument digital interface [14,15] for frequency and velocity of FT, and a goniometer [16], a three-dimensional (3D) motion and position measurement system [17–20] or image-based motion analyzer [21] for amplitude and/or velocity of FT. However,
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
these systems were not employed at the bedside or clinically. For daily clinical use, a measuring system should be low in cost, small in size, provide a large amount of information, and easy to use. No previous study has involved sequential motion analysis of index finger-to-thumb oppositions in clinical conditions, except for one previously reported by our group [3] and another using a potentiometer reported by Oliveira et al. [22]. In the present study, we developed an FT movement measurement system involving two tri-axial accelerometers, a pair of touch sensors, and a personal computer. This was easily applicable in a clinical situation and at the bedside to measure precisely various parameters such as rhythm, amplitude and velocity during FT performance. On the basis of our findings, we propose novel and objective parameters of FT movement in PD. 2. Methods 2.1. Finger tapping movement measurement system Fig. 1A shows the FT movement measurement system we developed. This system comprised two tri-axial accelerometers (PEA-350, Wacoh Corp., Japan), touch sensors, an AD converter (DAQCard-6036E, National Instruments Corp., Austin, USA) and a personal computer (ThinkPad T41, IBM). The accelerometer measured 20.5 12.5 5.0 mm and weighed 4.0 g, which was small and light enough to allow the FT movements to be performed naturally. The frequency bandwidth was 0.5–2000 Hz. The accelerometers were placed on the distal interphalangeal joint of the thumb and index finger by mounting them on fingerstalls. The touch sensors attached to the fingerstalls were made of thin stainless steel sheets, and designed to fit the ventral surface of the thumb and index finger. The touch sensors were used to indicate exactly when the thumb made contact with the index finger. The outputs of the accelerometers and the touch sensors were AD-converted using a sampling period of 0.1 ms and stored in a personal computer. 2.2. Parameters extracted from finger tapping movement Based on information from the outputs of the accelerometer and touch sensors, we calculated the rhythm, amplitude and velocity of the FT movement as shown in Fig. 1B: Fig. 1B(a) shows measured
441
acceleration, Fig. 1B(b) the output of the touch sensor, Fig. 1B(c) the velocity calculated by integrating the measured acceleration, and Fig. 1B(d) the distance between the thumb and the index finger, calculated by integrating the velocities. Our preliminary study [3] had shown that these various parameters obtained by this method are consistent with those obtained from a 3D measurement system (OPTOTRAK, Northern Digital Inc., Canada).
2.3. Measurement protocols and participating subjects Sixteen patients (mean age SD ¼ 64.9 9.8 yr, range 44–84 yr, mean disease duration SD ¼ 6.5 3.8 yr, total score of UPDRS motor examination in the ‘on’ state 27 13.6) fulfilling the diagnostic criteria for PD [1] and thirty-two healthy volunteers (HVs) (mean age SD ¼ 68.8 4.6 yr, range 45–79 yr) were recruited. All were right-handed. All patients with PD were receiving dopaminergic replacement therapy and tested in the ‘on’ state. We considered as exclusion criteria the presence of severe disabling dyskinesias, marked action tremor, severe cognitive impairment or marked bradykinesia according to clinical judgment based on UPDRS criteria. The participants were seated comfortably and moved their fingers parallel to the surface of the desk in order to avoid gravitational artifact in the principal (z) axis of the accelerometers [23] (Fig. 1A). Before measurements, the participants familiarized themselves with the devices and practiced FT for 10 s. Each participant was asked to move the thumb and the index finger as fast and as widely as possible for 60 s. With a sampling rate of 10 kHz, the signals from the accelerometers and the touch sensors were recorded and digitized by an IBM laptop computer with a 12-bit analog-digital (A/D) board and a data analyzer (LabVIEW, National Instruments Corp., Austin, USA). Subsequently, digitized data were analyzed with our original program written by MATLAB (The MathWorks Inc., Natick, Massachusetts, USA) to evaluate rhythm, velocity and amplitude. These experiments were executed with both the right and left hands, and all FT tests were videotaped. An independent neurologist evaluated the UPDRS FT score on the videotape to investigate its association with parameters obtained by this system.
Fig. 1. (A) System for measurement of finger tapping (FT) movement. (B) Measured acceleration and calculated velocity and amplitude during FT movement: (a) measured acceleration, (b) output of touch sensor, (c) calculated velocity, and (d) calculated amplitude. Ai indicates the time when the fingers make contact, and Bi indicates the time when the fingers separate, where i is the number of finger taps. Ti is defined as the single finger tapping interval (FTI) between the onset of finger tap and the onset of the next finger tap. Tci is defined as the single finger movement interval (FMI) between the onset and offset of finger tap, and Tmi is the single finger contact interval (FCI) between the offset and next onset of finger tap. Voi is the maximum opening velocity in a single FT movement. Vci is the maximum closing velocity in a single FT movement. Pi is the maximum amplitude during a single FT movement.
442
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
The study protocol was approved by the Osaka University Graduate School of Medicine – Ethic Committee. Informed consent was obtained from all the participants. 2.4. Analysis and statistics Data from both hands were combined in the analysis. Both mean and standard deviation (SD) of three parameters – FTI, FMI and FCI (see Fig. 1B) – signified the index of rhythm as the variation of FT coordination. The FT velocity was estimated from the maximum opening velocity (MoV) in opening phase and the maximum closing velocity (McV) in closing phase for each FT phase. The mean and SD of MoV and McV for 60 s were calculated. The FT amplitude was the difference between the displacement for the z-axis component of the thumb and the index finger. In addition, it was estimated from the maximum amplitude (MA) in each phase. The mean and SD of MA for 60 s were calculated. Moreover, the scale that evaluated the total distance of FT movement (TD) was defined as the sum total of MA for 60 s. The frequency was number of FT for 60 s (NFT). Mean and SD of the fourteen parameters mentioned above were calculated. In addition, each of the fourteen parameters was normalized. The radar chart displayed the differences in fourteen parameters between patients with PD and HVs. In order to clarify how these parameters were related to each other during FT performance, principal component analysis (PCA) was adopted. PCA is a technique for discovering or reducing the dimensionality of data, and can provide new meaningful variables for FT performance [24]. Without any distinction between HVs and patients with PD, fourteen parameters were analyzed and grouped by PCA. Simple logistic regression analysis was used when the response variable was a binary variable to extract important parameters from the fourteen that were available. The Akaike Information Criterion (AIC) is an index commonly used in a number of areas as an aid for choosing between competing models [25], lower values of the AIC index indicating the preferred model. The misclassification rate of each of the fourteen parameters was calculated. In order to clarify the relationship between each parameter and the UPDRS FT score, the boxplots of representative parameters are shown. All of HVs were included in the UPDRS FT score ‘‘0’’ by an independent neurologist. Spearman’s correlation coefficients were employed to investigate the relationship between UPDRS FT score and the representative parameters. 3. Results 3.1. Radar chart The results from the radar chart are shown in Fig. 2. In the velocity and amplitude component, there was an obvious difference between HVs and patients with PD. Particularly, mean of MoV and TD showed a marked difference between the two groups. In contrast, the rhythm component showed less of a difference than the velocity and amplitude component. In addition, NFT and mean of FMI showed no inter-group difference. As a whole, the radar chart for velocity, amplitude and rhythm revealed differences between patients with PD and HVs. 3.2. Principal component analysis Table 1 shows the contributory rates of the fourteen parameters. The first principal component accounted for 40.3% of the total variance of the original data. Seven parameters (Mean of MoV, Mean of McV, SD of MoV, SD of McV, Mean of MA, SD of MA, and TD) contributed to the first principal component. The second
Fig. 2. Radar chart of the fourteen standardized parameters. Each of the parameters is standardized. The chart displays the differences in the fourteen parameters between patients with PD and age-matched healthy volunteers. FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger tapping movement; NFT, number of finger tapping for 60 s.
principal component accounted for 27.4%. Four parameters (mean of FTI, mean of FMI, mean of FCI, and NFT) contributed to the second principal component. The third principal component accounted for 12.3%. The third principal component included parameters involved in the variation of interval and velocity. Among them, SD of FTI and SD of FMI were representative. The cumulative contributory rate from the first component to the third was 80.0%, which was enough to explain the fourteen parameters by these three components.
3.3. Simple logistic regression The results of simple logistic regression are shown in Table 2. It was evident that mean MoV and TD were useful parameters. Above all, this analysis revealed that mean MoV was the most prominent marker, with a misclassification rate/AIC of 15.6%/ 85.9, followed by TD with a misclassification rate/AIC of 18.8%/ 85.4.
Table 1 Contributory ratea of fourteen parameters in each principal component. Parameter
Component (contributory rate of the total) First (40.3)
Second (27.4)
Third (12.3)
Mean of FTI SD of FTI Mean of FMI SD of FMI Mean of FCI SD of FCI Mean of MoV SD of MoV Mean of McV SD of McV Mean of MA SD of MA TD NFT
0.1 3.4 0 3.0 0.5 0.9 15.6 9.0 15.1 10.9 14.6 11.1 15.4 0.2
23.1 6.6 15.1 5.8 14.6 8.8 0.3 0 0.1 0.2 1.2 1.4 0.4 22.2
5.1 25.9 1.8 19.9 5.1 8.0 0 11.9 0.4 8.2 1.0 7.9 0 4.8
FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger movement; NFT, number; and NFT, number of finger tapping 60 s. a All data are presented as percentages.
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444 Table 2 Simple logistic regression analysis for fourteen parameters. Parameter
AIC
Misclassification rate (%)
Mean of FTI SD of FTI Mean of FMI SD of FMI Mean of FCI SD of FCI Mean of MoV SD of MoV Mean of McV SD of McV Mean of MA SD of MA TD NFT
123.8 118.1 126.2 124.4 120.1 103.1 85.9 102.9 105.7 114.4 95.7 106.1 85.4 125.5
30.2 32.3 33.3 30.2 29.2 20.8 15.6 24.0 25.0 29.2 22.9 25.0 18.8 33.3
AIC, Akaike Information Criterion; FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger tapping movement; and NFT, number of finger tapping for 60 s.
3.4. Box-plot Box-plots of representative parameters (mean of MoV, TD, NFT and SD of FTI) from each component are shown in Fig. 3. From the results of PCA, mean of MoV and TD represented the first principal component, NFT represented the second, and SD of FTI represented the third. Mean of MoV and TD decreased in accordance with deterioration of the UPDRS FT score. SD of FTI increased in accordance with deterioration of the UPDRS FT score. These three parameters showed a close relationship with the severity of the
443
UPDRS FT score. However, NFT showed a poor relationship with the severity of the UPDRS FT score. Particularly, Spearman’s correlation coefficient between mean of MoV and UPDRS FT score was 0.59, which was the highest among the representative parameters. 4. Discussion For evaluation of the FT test, the UPDRS FT score has been widely adopted internationally [26]. This score is determined by eye, which sometimes results in error [5]. For example, although the velocity of FT movement in patients with PD is slightly higher with quite a small amplitude, so-called festination, than in HVs, the velocity in patients with festination was rather slower than in HVs when measured using this system (data not shown). Similar results have been reported by Konczak [17]. Therefore, a simple and objective method for assessing the FT test is required. Since the system we developed for this study is handy, simple, efficient and reproducible, it is the best one reported so far for evaluation of the FT test at the bedside. Fourteen parameters were collected from patients at the clinic in a short time, and then analyzed systematically. To recognize, at a glance, the difference between HVs and patients with PD, a radar chart (Fig. 2) for the fourteen parameters was useful. In particular, the parameters for velocity and amplitude revealed marked differences between the two groups. After successful measurement of so many parameters, we then had to consider the relationship of each parameter and its significance in FT performance. Therefore, we analyzed the parameters by PCA and simple logistic regression. PCA for the fourteen parameters showed that FT movement consisted of three components: mean and SD of amplitude and velocity as the first component, NFT and mean of FT interval as the
Fig. 3. Box-plots of representative parameters. (A) Mean of MoV, (B) TD, (C) NFT and (D) SD of FTI. Each box-plot of the representative parameter (mean of MoV, TD, NFT, SD of FTI) is shown in each rating stage of the UPDRS FT score. Mean of MoV, TD, and SD of FTI have close relationship with the severity of the UPDRS FT score. However, NFT has a poor relationship with the severity of the UPDRS FT score.
444
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
second component, and SD of FT intervals as the third component. Velocity- and amplitude-related parameters in the first principal component were valuable for the FT test. Among them, mean of MoV and TD with high contributory rates (Table 1), representative of the first principal component, were demonstrated to be the best two markers by simple logistic regression. These two markers were then further analyzed to investigate their relationship with UPDRS FT score (Fig. 3A and B), and it was shown that MoV was better than TD by Spearman’s correlation coefficient. NFT and FT intervals contribute to the second principal component, which is not useful for distinguishing HVs from patients with PD (Table 2) or for evaluation of PD severity (Fig. 3C). The third principal component consisted of variables involved in rhythm and velocity. SD of FTI, representative of the third principal component, was useful for the FT test (Fig. 3D), but less valuable than MoV and TD. Finally, we propose MoV as a novel and objective marker to assess PD severity. However, it does not still remain to be solved how these parameters change after treatments using drugs or deep brain stimulation (DBS) [14]. Since each component of FT performance seems to be related to different neural commands, parameters from the first and third components useful for the FT test in PD might respond to drug or DBS in different ways. Indeed, our preliminary study has shown that SD of FTI in the second component response was earlier than MoV after treatment with a dopamine agonist (Yokoe et al. unpublished data). This system may be applicable to other neurological disorders such as cerebellar ataxia [6]. As usage, it will be necessary to develop other parameters specifically involved in each symptom. Analysis of neurological symptoms using computer science should help to reveal the nature of each symptom, and as a result neurologic symptomatology might change to recognition of programming errors in the motor function of the brain. Acknowledgements This study was supported by the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (NIBIO). Appreciation is also extended to H. Ogawa and K. Hamada for experimental assistance. References [1] Calne DB, Snow BJ, Lee C. Criteria for diagnosing Parkinson’s disease. Ann Neurol 1992;32(Suppl.):125–7. [2] Kandori A, Yokoe M, Sakoda S, Abe K, Miyashita T, Oe H, et al. Quantitative magnetic detection of finger movements in patients with Parkinson’s disease. Neurosci Res 2004;49:253–60. [3] Okuno R, Yokoe M, Akazawa K, Abe K, Sakoda S. Finger taps movement acceleration measurement system for quantitative diagnosis of Parkinson’s disease. In: Proceedings of the twenty-eighth annual international conference of the IEEE Engineering in Medicine and Biology Society; 2006. p. 6623–6.
[4] Agostino R, Berardelli A, Curra A, Accornero N, Manfredi M. Clinical impairment of sequential finger movements in Parkinson’s disease. Mov Disord 1998;13:418–21. [5] Goetz CG, Stebbins GT. Assuring interrater reliability for the UPDRS motor section: utility of the UPDRS teaching tape. Mov Disord 2004;12:1453–6. [6] Shimoyama I, Ninchoji T, Uemura K. The finger-tapping test. A quantitative analysis. Arch Neurol 1990;47:681–4. [7] Homann CN, Suppan K, Wenzel K, Giovannoni G, Ivanic G, Horner S, et al. The Bradykinesia Akinesia Incoordination Test (BRAIN TEST), an objective and user-friendly means to evaluate patients with parkinsonism. Mov Disord 2000;15:641–7. [8] Giovannoni G, van Schalkwyk J, Fritz VU, Lees AJ. Bradykinesia akinesia incoordination test (BRAIN TEST): an objective computerized evaluation of upper limb motor function. J Neurol Neurosurg Psychiatr 1999;67:624–9. [9] Freeman JS, Cody FW, Schady W. The influence of external timing cues upon the rhythm of voluntary movements in Parkinson’s disease. J Neurol Neurosurg Psychiatr 1993;56:1078–84. [10] Nagasaki H, Itou H, Maruyama H, Hashizume K. Characteristic difficulty in rhythmic movement with aging and its relation to Parkinson’s disease. Exp Aging Res 1988;14:171–6. [11] Kennedy WR, Navarro X, Stewart NJ. Quantitation of the alternate movement rate in normal and diabetic subjects. Muscle Nerve 1991;14:1231–5. [12] Cousins MS, Corrow C, Finn M, Salamone JD. Temporal measures of human finger tapping: effects of age. Pharmacol Biochem Behav 1998;59:445–9. [13] Calautti C, Jones PS, Persaud N, Guincestre JY, Naccarato M, Warburton EA, et al. Quantification of index tapping regularity after stroke with tri-axial accelerometry. Brain Res Bull 2006;70:1–7. [14] Bronte-Stewart HM, Ding L, Alexander C, Zhou Y, Moore GP. Quantitative digitography (QDG): a sensitive measure of digital motor control in idiopathic Parkinson’s disease. Mov Disord 2000;15:36–47. [15] Taylor Tavares AL, Jefferis GS, Koop M, Hill BC, Hastie T, Heit G, et al. Quantitative measurements of alternating finger tapping in Parkinson’s disease correlate with UPDRS motor disability and reveal the improvement in fine motor control from medication and deep brain stimulation. Mov Disord 2005;20:1286–98. [16] Kimber TE, Tsai CS, Semmler J, Brophy BP, Thompson PD. Voluntary movement after pallidotomy in severe Parkinson’s disease. Brain 1999;122:895–906. [17] Konczak J, Ackermann H, Hertrich I, Spieker S, Dichgans J. Control of repetitive lip and finger movements in Parkinson’s disease: influence of external timing signals and simultaneous execution on motor performance. Mov Disord 1997;12:665–76. [18] Agostino R, Curra A, Giovannelli M, Mondugno N, Manfredi M, Berardelli A. Impairment of individual finger movements in Parkinson’s disease. Mov Disord 2003;18:560–5. [19] Frischer M. Voluntary vs autonomous control of repetitive finger-tapping in a patient with Parkinson’s disease. Neuropsychologia 1989;27:1261–6. [20] Keresztenyi Z, Valkovic P, Eggert T, Steude U, Hermsdo¨rfer J, Laczko J, et al. The time course of the return of upper limb bradykinesia after cessation of subthalamic stimulation in Parkinson’s disease. Parkinsonism Relat Disord 2007;13:438–42. [21] Jobba´gy A´ Harcos P, Karoly R, Fazekas G. Analysis of finger-tapping movement. J Neurosci Methods 2005;141:29–39. [22] Oliveira RM, Gurd JM, Nixon P, Marshall JC, Passingham RE. Hypometria in Parkinson’s disease: automatic versus controlled processing. Mov Disord 1998;13:422–7. [23] Elble JR. Gravitational artifact in accelerometric measurements of tremor. Clin Neurophysiol 2005;116:1638–43. [24] Boonstra TW, Daffertshofer A, Peper CE, Beek PJ. Amplitude and phase dynamics associated with acoustically paced finger tapping. Brain Res 2006:60–9. [25] Akaike H. A new look at the statistical model identification. IEEE Trans Automat Control 1974:716–23. [26] Fahn S, Elton RL, members of the UPDRS Development Committee. Unified Parkinson’s disease rating scale. In: Fahn S, Marsden CD, Calne DB, Goldstein M, editors. Recent developments in Parkinson’s disease, vol. 2. Florham Park, NJ: MacMillan Healthcare Information; 1987. p. 153–63.
Contents lists available at ScienceDirect
Parkinsonism and Related Disorders journal homepage: www.elsevier.com/locate/parkreldis
Opening velocity, a novel parameter, for finger tapping test in patients with Parkinson’s diseaseq M. Yokoe a, e, R. Okuno d, T. Hamasaki b, e, Y. Kurachi c, e, K. Akazawa f, S. Sakoda a, e, * a
Department of Neurology, Osaka University Graduate School of Medicine, Osaka, Japan Department of Medical Statistics, Osaka University Graduate School of Medicine, Osaka, Japan c Department of Pharmacology II, Osaka University Graduate School of Medicine, Osaka, Japan d Department of Electrical and Electronics Engineering, Faculty of Engineering, Setsunan University, Osaka, Japan e The Center for Advanced Medical Engineering and Informatics, Osaka University, Osaka, Japan f Department of Biomedical Engineering, Osaka Institute of Technology, Osaka, Japan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 24 June 2008 Received in revised form 19 August 2008 Accepted 1 November 2008
Objectives: A new system consisting of an accelerometer and touch sensor was developed to find objective parameters for the finger tapping (FT) test in patients with Parkinson’s disease (PD). Methods: We recruited sixteen patients with PD and thirty-two age-matched healthy volunteers (HVs). By using this new system, various parameters related to velocity, amplitude, rhythm and number in the FT test were measured in patients with PD and examined in comparison with those of HVs on the basis of the Unified Parkinson’s Disease Rating Scale (UPDRS) FT score. Results: The new system allowed us to measure fourteen parameters of FT movement very easily, and a radar chart showed obvious differences in most of these parameters between HVs and patients with PD. Principal component analysis showed that fourteen parameters were classified into three components: (1) both mean and standard deviation (SD) of both amplitude and velocity, (2) number of FT for 60 s and mean FT interval, and (3) SD of FT interval. The first (velocity- and amplitude-related parameters) and third (rhythm-related parameters) components contributed to discrimination of PD from HVs. Maximum opening velocity (MoV) was the best of these parameters because of its sensitivity and association with the UPDRS FT score. Conclusions: A novel system for the FT test, which is compact, simple and efficient, has been developed. Velocity- and amplitude-related parameters were indicated to be valuable for evaluation of the FT test in patients with PD. In particular, we first propose that MoV is a novel marker for the FT test. Ó 2008 Elsevier Ltd. All rights reserved.
Keywords: Parkinson’s disease Finger tapping Accelerometer Principal component analysis
1. Introduction Although four clinical symptoms are well known in Parkinson’s disease (PD) – tremor, rigidity, bradykinesia/akinesia and absence of postural reflexes – the nature of each symptom [1] and how symptoms such as rigidity affect movement in patients with PD are still unclear. From a scientific viewpoint to understand the nature of PD movement, it is important to analyze movement from the aspects of rhythm, velocity and amplitude. From a clinical viewpoint, an objective system for assessment of clinical symptoms is required because quantification and consistency in the assessment of PD severity would be of considerable help in clarifying the q The review of this paper was entirely handled by an Associate Editor, Prof. Eng-King Tan. * Corresponding author at: Department of Neurology, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka, Japan. Tel.: þ81 6 6879 3571; fax: þ81 6 6879 3579. E-mail address: [email protected] (S. Sakoda). 1353-8020/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.parkreldis.2008.11.003
efficacy of various drugs. We have focused on the UPDRS finger tapping (FT) score and reported a prototype method for its objective and quantitative evaluation [2,3] because FT as a parameter of upper limb motor function [4] has been widely used to assess motor impairment in patients with neurological or neuropsychological disorders. Accurate scoring and interpretation of results require experience, and even experienced neurologists are unable to detect subtle changes. Goetz and colleagues reported that the UPDRS FT score was one of the most difficult items to assess [5]. Accuracy and reproducibility in the quantitative analysis of FT performance are indispensable for evaluating the severity of PD. The FT test has been analyzed by various methods including the use of a computer keyboard [6–8], a metal plate [9–11], a telegraph key [12] or a triaxial accelerometer [13] for frequency and/or rhythm of FT, a musical instrument digital interface [14,15] for frequency and velocity of FT, and a goniometer [16], a three-dimensional (3D) motion and position measurement system [17–20] or image-based motion analyzer [21] for amplitude and/or velocity of FT. However,
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
these systems were not employed at the bedside or clinically. For daily clinical use, a measuring system should be low in cost, small in size, provide a large amount of information, and easy to use. No previous study has involved sequential motion analysis of index finger-to-thumb oppositions in clinical conditions, except for one previously reported by our group [3] and another using a potentiometer reported by Oliveira et al. [22]. In the present study, we developed an FT movement measurement system involving two tri-axial accelerometers, a pair of touch sensors, and a personal computer. This was easily applicable in a clinical situation and at the bedside to measure precisely various parameters such as rhythm, amplitude and velocity during FT performance. On the basis of our findings, we propose novel and objective parameters of FT movement in PD. 2. Methods 2.1. Finger tapping movement measurement system Fig. 1A shows the FT movement measurement system we developed. This system comprised two tri-axial accelerometers (PEA-350, Wacoh Corp., Japan), touch sensors, an AD converter (DAQCard-6036E, National Instruments Corp., Austin, USA) and a personal computer (ThinkPad T41, IBM). The accelerometer measured 20.5 12.5 5.0 mm and weighed 4.0 g, which was small and light enough to allow the FT movements to be performed naturally. The frequency bandwidth was 0.5–2000 Hz. The accelerometers were placed on the distal interphalangeal joint of the thumb and index finger by mounting them on fingerstalls. The touch sensors attached to the fingerstalls were made of thin stainless steel sheets, and designed to fit the ventral surface of the thumb and index finger. The touch sensors were used to indicate exactly when the thumb made contact with the index finger. The outputs of the accelerometers and the touch sensors were AD-converted using a sampling period of 0.1 ms and stored in a personal computer. 2.2. Parameters extracted from finger tapping movement Based on information from the outputs of the accelerometer and touch sensors, we calculated the rhythm, amplitude and velocity of the FT movement as shown in Fig. 1B: Fig. 1B(a) shows measured
441
acceleration, Fig. 1B(b) the output of the touch sensor, Fig. 1B(c) the velocity calculated by integrating the measured acceleration, and Fig. 1B(d) the distance between the thumb and the index finger, calculated by integrating the velocities. Our preliminary study [3] had shown that these various parameters obtained by this method are consistent with those obtained from a 3D measurement system (OPTOTRAK, Northern Digital Inc., Canada).
2.3. Measurement protocols and participating subjects Sixteen patients (mean age SD ¼ 64.9 9.8 yr, range 44–84 yr, mean disease duration SD ¼ 6.5 3.8 yr, total score of UPDRS motor examination in the ‘on’ state 27 13.6) fulfilling the diagnostic criteria for PD [1] and thirty-two healthy volunteers (HVs) (mean age SD ¼ 68.8 4.6 yr, range 45–79 yr) were recruited. All were right-handed. All patients with PD were receiving dopaminergic replacement therapy and tested in the ‘on’ state. We considered as exclusion criteria the presence of severe disabling dyskinesias, marked action tremor, severe cognitive impairment or marked bradykinesia according to clinical judgment based on UPDRS criteria. The participants were seated comfortably and moved their fingers parallel to the surface of the desk in order to avoid gravitational artifact in the principal (z) axis of the accelerometers [23] (Fig. 1A). Before measurements, the participants familiarized themselves with the devices and practiced FT for 10 s. Each participant was asked to move the thumb and the index finger as fast and as widely as possible for 60 s. With a sampling rate of 10 kHz, the signals from the accelerometers and the touch sensors were recorded and digitized by an IBM laptop computer with a 12-bit analog-digital (A/D) board and a data analyzer (LabVIEW, National Instruments Corp., Austin, USA). Subsequently, digitized data were analyzed with our original program written by MATLAB (The MathWorks Inc., Natick, Massachusetts, USA) to evaluate rhythm, velocity and amplitude. These experiments were executed with both the right and left hands, and all FT tests were videotaped. An independent neurologist evaluated the UPDRS FT score on the videotape to investigate its association with parameters obtained by this system.
Fig. 1. (A) System for measurement of finger tapping (FT) movement. (B) Measured acceleration and calculated velocity and amplitude during FT movement: (a) measured acceleration, (b) output of touch sensor, (c) calculated velocity, and (d) calculated amplitude. Ai indicates the time when the fingers make contact, and Bi indicates the time when the fingers separate, where i is the number of finger taps. Ti is defined as the single finger tapping interval (FTI) between the onset of finger tap and the onset of the next finger tap. Tci is defined as the single finger movement interval (FMI) between the onset and offset of finger tap, and Tmi is the single finger contact interval (FCI) between the offset and next onset of finger tap. Voi is the maximum opening velocity in a single FT movement. Vci is the maximum closing velocity in a single FT movement. Pi is the maximum amplitude during a single FT movement.
442
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
The study protocol was approved by the Osaka University Graduate School of Medicine – Ethic Committee. Informed consent was obtained from all the participants. 2.4. Analysis and statistics Data from both hands were combined in the analysis. Both mean and standard deviation (SD) of three parameters – FTI, FMI and FCI (see Fig. 1B) – signified the index of rhythm as the variation of FT coordination. The FT velocity was estimated from the maximum opening velocity (MoV) in opening phase and the maximum closing velocity (McV) in closing phase for each FT phase. The mean and SD of MoV and McV for 60 s were calculated. The FT amplitude was the difference between the displacement for the z-axis component of the thumb and the index finger. In addition, it was estimated from the maximum amplitude (MA) in each phase. The mean and SD of MA for 60 s were calculated. Moreover, the scale that evaluated the total distance of FT movement (TD) was defined as the sum total of MA for 60 s. The frequency was number of FT for 60 s (NFT). Mean and SD of the fourteen parameters mentioned above were calculated. In addition, each of the fourteen parameters was normalized. The radar chart displayed the differences in fourteen parameters between patients with PD and HVs. In order to clarify how these parameters were related to each other during FT performance, principal component analysis (PCA) was adopted. PCA is a technique for discovering or reducing the dimensionality of data, and can provide new meaningful variables for FT performance [24]. Without any distinction between HVs and patients with PD, fourteen parameters were analyzed and grouped by PCA. Simple logistic regression analysis was used when the response variable was a binary variable to extract important parameters from the fourteen that were available. The Akaike Information Criterion (AIC) is an index commonly used in a number of areas as an aid for choosing between competing models [25], lower values of the AIC index indicating the preferred model. The misclassification rate of each of the fourteen parameters was calculated. In order to clarify the relationship between each parameter and the UPDRS FT score, the boxplots of representative parameters are shown. All of HVs were included in the UPDRS FT score ‘‘0’’ by an independent neurologist. Spearman’s correlation coefficients were employed to investigate the relationship between UPDRS FT score and the representative parameters. 3. Results 3.1. Radar chart The results from the radar chart are shown in Fig. 2. In the velocity and amplitude component, there was an obvious difference between HVs and patients with PD. Particularly, mean of MoV and TD showed a marked difference between the two groups. In contrast, the rhythm component showed less of a difference than the velocity and amplitude component. In addition, NFT and mean of FMI showed no inter-group difference. As a whole, the radar chart for velocity, amplitude and rhythm revealed differences between patients with PD and HVs. 3.2. Principal component analysis Table 1 shows the contributory rates of the fourteen parameters. The first principal component accounted for 40.3% of the total variance of the original data. Seven parameters (Mean of MoV, Mean of McV, SD of MoV, SD of McV, Mean of MA, SD of MA, and TD) contributed to the first principal component. The second
Fig. 2. Radar chart of the fourteen standardized parameters. Each of the parameters is standardized. The chart displays the differences in the fourteen parameters between patients with PD and age-matched healthy volunteers. FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger tapping movement; NFT, number of finger tapping for 60 s.
principal component accounted for 27.4%. Four parameters (mean of FTI, mean of FMI, mean of FCI, and NFT) contributed to the second principal component. The third principal component accounted for 12.3%. The third principal component included parameters involved in the variation of interval and velocity. Among them, SD of FTI and SD of FMI were representative. The cumulative contributory rate from the first component to the third was 80.0%, which was enough to explain the fourteen parameters by these three components.
3.3. Simple logistic regression The results of simple logistic regression are shown in Table 2. It was evident that mean MoV and TD were useful parameters. Above all, this analysis revealed that mean MoV was the most prominent marker, with a misclassification rate/AIC of 15.6%/ 85.9, followed by TD with a misclassification rate/AIC of 18.8%/ 85.4.
Table 1 Contributory ratea of fourteen parameters in each principal component. Parameter
Component (contributory rate of the total) First (40.3)
Second (27.4)
Third (12.3)
Mean of FTI SD of FTI Mean of FMI SD of FMI Mean of FCI SD of FCI Mean of MoV SD of MoV Mean of McV SD of McV Mean of MA SD of MA TD NFT
0.1 3.4 0 3.0 0.5 0.9 15.6 9.0 15.1 10.9 14.6 11.1 15.4 0.2
23.1 6.6 15.1 5.8 14.6 8.8 0.3 0 0.1 0.2 1.2 1.4 0.4 22.2
5.1 25.9 1.8 19.9 5.1 8.0 0 11.9 0.4 8.2 1.0 7.9 0 4.8
FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger movement; NFT, number; and NFT, number of finger tapping 60 s. a All data are presented as percentages.
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444 Table 2 Simple logistic regression analysis for fourteen parameters. Parameter
AIC
Misclassification rate (%)
Mean of FTI SD of FTI Mean of FMI SD of FMI Mean of FCI SD of FCI Mean of MoV SD of MoV Mean of McV SD of McV Mean of MA SD of MA TD NFT
123.8 118.1 126.2 124.4 120.1 103.1 85.9 102.9 105.7 114.4 95.7 106.1 85.4 125.5
30.2 32.3 33.3 30.2 29.2 20.8 15.6 24.0 25.0 29.2 22.9 25.0 18.8 33.3
AIC, Akaike Information Criterion; FTI, single finger tapping interval; SD, standard deviation; FMI, single finger movement interval; FCI, single finger contact interval; MoV, maximum opening velocity; McV, maximum closing velocity; MA, maximum amplitude; TD, total distance of finger tapping movement; and NFT, number of finger tapping for 60 s.
3.4. Box-plot Box-plots of representative parameters (mean of MoV, TD, NFT and SD of FTI) from each component are shown in Fig. 3. From the results of PCA, mean of MoV and TD represented the first principal component, NFT represented the second, and SD of FTI represented the third. Mean of MoV and TD decreased in accordance with deterioration of the UPDRS FT score. SD of FTI increased in accordance with deterioration of the UPDRS FT score. These three parameters showed a close relationship with the severity of the
443
UPDRS FT score. However, NFT showed a poor relationship with the severity of the UPDRS FT score. Particularly, Spearman’s correlation coefficient between mean of MoV and UPDRS FT score was 0.59, which was the highest among the representative parameters. 4. Discussion For evaluation of the FT test, the UPDRS FT score has been widely adopted internationally [26]. This score is determined by eye, which sometimes results in error [5]. For example, although the velocity of FT movement in patients with PD is slightly higher with quite a small amplitude, so-called festination, than in HVs, the velocity in patients with festination was rather slower than in HVs when measured using this system (data not shown). Similar results have been reported by Konczak [17]. Therefore, a simple and objective method for assessing the FT test is required. Since the system we developed for this study is handy, simple, efficient and reproducible, it is the best one reported so far for evaluation of the FT test at the bedside. Fourteen parameters were collected from patients at the clinic in a short time, and then analyzed systematically. To recognize, at a glance, the difference between HVs and patients with PD, a radar chart (Fig. 2) for the fourteen parameters was useful. In particular, the parameters for velocity and amplitude revealed marked differences between the two groups. After successful measurement of so many parameters, we then had to consider the relationship of each parameter and its significance in FT performance. Therefore, we analyzed the parameters by PCA and simple logistic regression. PCA for the fourteen parameters showed that FT movement consisted of three components: mean and SD of amplitude and velocity as the first component, NFT and mean of FT interval as the
Fig. 3. Box-plots of representative parameters. (A) Mean of MoV, (B) TD, (C) NFT and (D) SD of FTI. Each box-plot of the representative parameter (mean of MoV, TD, NFT, SD of FTI) is shown in each rating stage of the UPDRS FT score. Mean of MoV, TD, and SD of FTI have close relationship with the severity of the UPDRS FT score. However, NFT has a poor relationship with the severity of the UPDRS FT score.
444
M. Yokoe et al. / Parkinsonism and Related Disorders 15 (2009) 440–444
second component, and SD of FT intervals as the third component. Velocity- and amplitude-related parameters in the first principal component were valuable for the FT test. Among them, mean of MoV and TD with high contributory rates (Table 1), representative of the first principal component, were demonstrated to be the best two markers by simple logistic regression. These two markers were then further analyzed to investigate their relationship with UPDRS FT score (Fig. 3A and B), and it was shown that MoV was better than TD by Spearman’s correlation coefficient. NFT and FT intervals contribute to the second principal component, which is not useful for distinguishing HVs from patients with PD (Table 2) or for evaluation of PD severity (Fig. 3C). The third principal component consisted of variables involved in rhythm and velocity. SD of FTI, representative of the third principal component, was useful for the FT test (Fig. 3D), but less valuable than MoV and TD. Finally, we propose MoV as a novel and objective marker to assess PD severity. However, it does not still remain to be solved how these parameters change after treatments using drugs or deep brain stimulation (DBS) [14]. Since each component of FT performance seems to be related to different neural commands, parameters from the first and third components useful for the FT test in PD might respond to drug or DBS in different ways. Indeed, our preliminary study has shown that SD of FTI in the second component response was earlier than MoV after treatment with a dopamine agonist (Yokoe et al. unpublished data). This system may be applicable to other neurological disorders such as cerebellar ataxia [6]. As usage, it will be necessary to develop other parameters specifically involved in each symptom. Analysis of neurological symptoms using computer science should help to reveal the nature of each symptom, and as a result neurologic symptomatology might change to recognition of programming errors in the motor function of the brain. Acknowledgements This study was supported by the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (NIBIO). Appreciation is also extended to H. Ogawa and K. Hamada for experimental assistance. References [1] Calne DB, Snow BJ, Lee C. Criteria for diagnosing Parkinson’s disease. Ann Neurol 1992;32(Suppl.):125–7. [2] Kandori A, Yokoe M, Sakoda S, Abe K, Miyashita T, Oe H, et al. Quantitative magnetic detection of finger movements in patients with Parkinson’s disease. Neurosci Res 2004;49:253–60. [3] Okuno R, Yokoe M, Akazawa K, Abe K, Sakoda S. Finger taps movement acceleration measurement system for quantitative diagnosis of Parkinson’s disease. In: Proceedings of the twenty-eighth annual international conference of the IEEE Engineering in Medicine and Biology Society; 2006. p. 6623–6.
[4] Agostino R, Berardelli A, Curra A, Accornero N, Manfredi M. Clinical impairment of sequential finger movements in Parkinson’s disease. Mov Disord 1998;13:418–21. [5] Goetz CG, Stebbins GT. Assuring interrater reliability for the UPDRS motor section: utility of the UPDRS teaching tape. Mov Disord 2004;12:1453–6. [6] Shimoyama I, Ninchoji T, Uemura K. The finger-tapping test. A quantitative analysis. Arch Neurol 1990;47:681–4. [7] Homann CN, Suppan K, Wenzel K, Giovannoni G, Ivanic G, Horner S, et al. The Bradykinesia Akinesia Incoordination Test (BRAIN TEST), an objective and user-friendly means to evaluate patients with parkinsonism. Mov Disord 2000;15:641–7. [8] Giovannoni G, van Schalkwyk J, Fritz VU, Lees AJ. Bradykinesia akinesia incoordination test (BRAIN TEST): an objective computerized evaluation of upper limb motor function. J Neurol Neurosurg Psychiatr 1999;67:624–9. [9] Freeman JS, Cody FW, Schady W. The influence of external timing cues upon the rhythm of voluntary movements in Parkinson’s disease. J Neurol Neurosurg Psychiatr 1993;56:1078–84. [10] Nagasaki H, Itou H, Maruyama H, Hashizume K. Characteristic difficulty in rhythmic movement with aging and its relation to Parkinson’s disease. Exp Aging Res 1988;14:171–6. [11] Kennedy WR, Navarro X, Stewart NJ. Quantitation of the alternate movement rate in normal and diabetic subjects. Muscle Nerve 1991;14:1231–5. [12] Cousins MS, Corrow C, Finn M, Salamone JD. Temporal measures of human finger tapping: effects of age. Pharmacol Biochem Behav 1998;59:445–9. [13] Calautti C, Jones PS, Persaud N, Guincestre JY, Naccarato M, Warburton EA, et al. Quantification of index tapping regularity after stroke with tri-axial accelerometry. Brain Res Bull 2006;70:1–7. [14] Bronte-Stewart HM, Ding L, Alexander C, Zhou Y, Moore GP. Quantitative digitography (QDG): a sensitive measure of digital motor control in idiopathic Parkinson’s disease. Mov Disord 2000;15:36–47. [15] Taylor Tavares AL, Jefferis GS, Koop M, Hill BC, Hastie T, Heit G, et al. Quantitative measurements of alternating finger tapping in Parkinson’s disease correlate with UPDRS motor disability and reveal the improvement in fine motor control from medication and deep brain stimulation. Mov Disord 2005;20:1286–98. [16] Kimber TE, Tsai CS, Semmler J, Brophy BP, Thompson PD. Voluntary movement after pallidotomy in severe Parkinson’s disease. Brain 1999;122:895–906. [17] Konczak J, Ackermann H, Hertrich I, Spieker S, Dichgans J. Control of repetitive lip and finger movements in Parkinson’s disease: influence of external timing signals and simultaneous execution on motor performance. Mov Disord 1997;12:665–76. [18] Agostino R, Curra A, Giovannelli M, Mondugno N, Manfredi M, Berardelli A. Impairment of individual finger movements in Parkinson’s disease. Mov Disord 2003;18:560–5. [19] Frischer M. Voluntary vs autonomous control of repetitive finger-tapping in a patient with Parkinson’s disease. Neuropsychologia 1989;27:1261–6. [20] Keresztenyi Z, Valkovic P, Eggert T, Steude U, Hermsdo¨rfer J, Laczko J, et al. The time course of the return of upper limb bradykinesia after cessation of subthalamic stimulation in Parkinson’s disease. Parkinsonism Relat Disord 2007;13:438–42. [21] Jobba´gy A´ Harcos P, Karoly R, Fazekas G. Analysis of finger-tapping movement. J Neurosci Methods 2005;141:29–39. [22] Oliveira RM, Gurd JM, Nixon P, Marshall JC, Passingham RE. Hypometria in Parkinson’s disease: automatic versus controlled processing. Mov Disord 1998;13:422–7. [23] Elble JR. Gravitational artifact in accelerometric measurements of tremor. Clin Neurophysiol 2005;116:1638–43. [24] Boonstra TW, Daffertshofer A, Peper CE, Beek PJ. Amplitude and phase dynamics associated with acoustically paced finger tapping. Brain Res 2006:60–9. [25] Akaike H. A new look at the statistical model identification. IEEE Trans Automat Control 1974:716–23. [26] Fahn S, Elton RL, members of the UPDRS Development Committee. Unified Parkinson’s disease rating scale. In: Fahn S, Marsden CD, Calne DB, Goldstein M, editors. Recent developments in Parkinson’s disease, vol. 2. Florham Park, NJ: MacMillan Healthcare Information; 1987. p. 153–63.