A Novel Glove Monitoring System Used to Quantify ... - IEEE Xplore

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

3193

A Novel Glove Monitoring System Used to Quantify Neurological Symptoms During Deep-Brain Stimulation Surgery Houde Dai, Bernward Otten, Jan Hinnerk Mehrkens, L. T. D’Angelo, and Tim C. Lueth

Abstract— Deep brain stimulation (DBS) surgery is most effective in reducing the symptoms of Parkinson’s disease and essential tremor. At present, there is no designated instrumental method for measuring the immediate effects of DBS. This paper presents the concept of a glove monitoring system for DBS. With the benefits of microelectromechanical systems, inertial measurement unit, and force sensitive resistor (FSR), the system is portable and can be integrated into a textile glove. Tremors, bradykinesia, and rigidity assessments are performed by the system. Several test tasks are chosen to be performed during DBS surgery to evaluate the electrode’s position and stimulation intensity. Each quantified symptom severity of the patient is added to a list shown in the graphical user interface for comparison. Comparative experiments between the prototype and an electromagnetic motion tracking system were presented. The FSR boxes were validated with weights. Experimental results show that this system is reliable for tremor amplitude determination, movement angles measurement, and resistance measurement to a passive movement. In addition, it can be found that inconsistent tremor movements have an influence on the tremor amplitude calculation done with power spectral density estimation. Index Terms— MEMS IMU, glove monitoring system, parkinson’s disease quantitative assessment, tremor, bradykinesia, rigidity, reliability testing.

I. I NTRODUCTION

D

EEP-BRAIN stimulation (DBS) is a surgical procedure used to treat neurological symptoms, mostly for Parkinson’s disease (PD) and essential tremor (ET). Before the procedure, MRI or CT scanning is used to identify and locate the target area of each electrode. About 5 to 10 points inside the target area are tested to evaluate the electrode target position and stimulation intensity [1]. PD is a progressive movement disorder. The symptoms of PD occur mainly in the patient’s hands, feet, and head.

Manuscript received March 12, 2013; revised June 5, 2013; accepted June 20, 2013. Date of publication July 9, 2013; date of current version July 30, 2013. This work was supported by the German Federal Ministry of Economics and Technology under Project KF2520905KJ2. The associate editor coordinating the review of this paper and approving it for publication was Prof. Aime Lay-Ekuakille. H. Dai, B. Otten, L. T. D’Angelo, and T. C. Lueth are with the Institute of Micro Technology and Medical Device Technology, Technische Universität München, Munich 85748, Germany (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). J. H. Mehrkens is with the Department of Neurology of Klinikum Großhadern, Ludwig-Maximilian University, München D-81377, Germany (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JSEN.2013.2271775

Three primary symptoms are tremors (oscillatory movement), bradykinesia (slow movement), and rigidity (increased muscle tone). All these symptoms can be assessed through several hand or elbow motion tasks. The tremor associated with PD has a characteristic appearance. It is mostly in the form of a rhythmic back-and-forth motion in the range of 3.5 to 7.5 Hz, which is called a “pill rolling” tremor. Tremors are classified into three primary types: rest tremor, postural tremor, and action tremor [2]. Resting tremors might happen when a body part is relaxed and completed supported against gravity, for example, when lying in bed. Postural tremors, which are the characteristic of ET, occur while the body part is voluntarily maintaining a position against gravity. Action tremors happen during a voluntary contraction of a muscle, for example, trying to hold a cup. It is uncommon to have a pure rest tremor. More common is the combination of rest, postural, and action tremors. Rest tremors are inhibited during movement and may reoccur during a postural tremor task. A mild form of action tremor is present in almost every PD patient and can be detected easily by analyzing slow flexion and extension movements. The action tremor score is not associated with age or disease duration [3]. Bradykinesia is a feature of basal ganglia disorders. It involves difficulties with planning, beginning, and executing movement and with performing sequential and simultaneous tasks. It often initially manifestasts as slowness in performing activities. Rigidity means increased muscle tone, which is defined as a resistance to a passive movement [4]. During DBS surgery, surface electromyography (sEMG) or intraoperative microelectrode recording (MER) [35] are usually used to assess the effects of surgery. However, EMG and MER are spiky impulse-like waveforms and the information about brain disorders is found in the morphology of an impulse chain. Impulse-like signals are hard to analyze using traditional amplitude and spectral methods [5]–[7]. Nevertheless, neurosurgeons need to know the exact position and stimulation intensity of the deep-stimulation electrode to achieve the best effect. According to the requirements of neurosurgeons and movement disorder neurologists, the goal of this project is to develop a glove monitoring system (GMS) to quantify hand tremors, bradykinesia, and rigidity in PD during DBS surgery.

1530-437X © 2013 IEEE

3194

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

II. S TATE OF THE A RT AND TASK D ESCRIPTION A. State of the Art in Hand Motion Monitoring Systems At present, there are some movement disorder assessment systems available in the market or for research. The Kinesia™ system by Great Lakes NeuroTechnologies is used to capture three-dimensional motion with a 3-axis gyroscope and 3-axis accelerometer on the upside of a finger, together with two channels EMG [8]. It is approved by the United States Food and Drug Administration (FDA). The Motus Movement Monitor by Motus Bioengineering Inc., only with a 3-axis gyroscope on the palm’s upside, is also used to assess tremors and bradykinesia in PD. Salarian et al. [9] presented a continuous hand motion monitor used to assess tremors and bradykinesia with a 3-axis gyroscope on the wrist. Niazmand et al. presented a smart glove [10] for the quantitative evaluation of PD with accelerometers, force sensor, and contact sensors based on a textile glove [11], [12]. All of the above mentioned systems are not for surgical application. They cannot assess all three primary PD symptoms. Most studies in PD assessment focused on the analysis of tremors [13]–[15]. Previous rigidity assessment systems are fixed in a place and mostly with large-scale structure [16], [17]. Recently MEMS (Micro-Electro-Mechanical Systems) technology has been greatly progressing. The integration of a 6-axis MEMS inertial measurement unit (IMU) (3-axis accelerometer and 3-axis gyroscope) in a small sensor chip offers miniaturization at low power. The assessment system for parkinsonian motor symptoms with new IMU can get better results because of its higher resolution and embedded algorithms. Accelerometer and gyroscope sensor data correlates strongly with the Unified Parkinson’s Disease Rating Scale (UPDRS) tremor scores [8]. The angular displacement and angular velocity during bradykinesia and rigidity assessment tasks can be acquired with gyroscopes [18]. The angular displacement and angular velocity represent the rhythm and amplitude, which are the variables of UPDRS bradykinesia and rigidity score models [19]. B. Challenges and Task Description Currently there is no system available for assessing all three primary PD symptoms. Therefore, it is important to realize three assessment tasks in one system. The device has to be realized as a pervasive device, based on a wired activity recorder and transceiver [20], [21]. Each subscale of the UPDRS is from 0 to 4, where 0 = normal, 1 = slight, 2 = mild, 3 = moderate and 4 = severe. Most of the previous research only compared the outputs of their systems to this 5-point rating scale. During DBS, a higher resolution for all parameters is needed. Thus, the resolution of tremor amplitude is set to 0.01 in this system [22]. For DBS electrode positioning, only the tremor amplitude is required, because the tremor frequency does not change for a PD patient For power spectral density (PSD) estimation, peak power means the power estimation around the dominant

TABLE I PARAMETERS AND T HEIR R EQUIRED A CCURACY

frequency in the power spectrum of a sensor signal According to the research conducted by Giuffrida et al. [23], for the rest tremor and postural tremor in PD, the logarithm of the peak powers summation of both power spectrums of accelerometer and gyroscope data had the highest correlation with UPDRS scores (coefficient of determination r 2 ≈ 0.9). For the action tremor in PD, the root-mean-square (RMS) sum of both gyroscope and accelerometer data had the highest correlation with UPDRS (r 2 = 0.69) [24] In this system, the logarithm of the peak powers summation of the IMU sensor signals is regarded as the tremor amplitude. Hand grasping is easier to perform during DBS. The signals of healthy people have a consistent amplitude and frequency, thus appearing sinusoidal. On the other hand, patients with severe bradykinesia have an inconsistent amplitude and frequency. Unlike for tremors, peak power during hand grasping is not correlated with the clinician UPDRS score. Instead, the mean value and standard deviation (SD) of hand grasping ranges (|α| and σ|α| ), where α is the peak-to-peak values of the hand grasping cycles, can be used as the parameters of bradykinesia. However, the correlation between these two values and UPDRS scores needs further study [25]. In the clinic, rigidity is assessed by a neurologist who moves the subject’s limb, scoring the result according to UPDRS. However, rigidity scores for an individual patient may vary depending on the examiner. Such scales are susceptible to problems of sensitivity and reliability. Mechanical impedance, which means the magnitude of the vector sum of elastic stiffness and viscous stiffness, is nonlinearly related to UPDRS rigidity ratings [26]. The tremor amplitudes, which are regressed from the peak powers of the IMU signals, can be gauged as the clinician ratings. However, the parameters of bradykinesia and rigidity cannot be normalized directly as UPDRS ratings at present. For calculating the accuracy settings in Table I, the parameters obtained from the glove monitoring systems with IMU and force sensors must be compared to the judgments of doctors according to UPDRS. The parameters should meet the accuracy setting as Table I [8], [27], [28]. In Table I, r 2 is the coefficient of determination and RMSE is the root-meansquare-error (RMSE). III. S YSTEM C ONCEPT According to the literature and the requirements of neurosurgeons, the system concept of a portable parkinsonian motor symptom assessment system was defined as follows.

DAI et al.: NOVEL GLOVE MONITORING SYSTEM USED TO QUANTIFY NEUROLOGICAL SYMPTOMS

(a)

(b)

3195

(c) (a)

Fig. 3. (d)

(b)

The force sensor housing box: (a) structure; (b) prototype.

(e)

Fig. 1. Test tasks relevant for DBS: (a) rest tremor; (b) postural tremor; (c) action tremor; (d) bradykinesia task (hand whole-hand grasping); (e) rigidity task (elbow flexion and extension).

Fig. 4.

(a)

Circuit in a force sensor box.

(b)

Fig. 2. (a) system diagram of the glove monitoring system; (b) rigidity assessment cuff. The components are: 1: IMU circuit board; 2: command module; 3: graphical user interface; 4: USB cable; 5: textile glove; 6: rigidity assessment cuff.

A. Test Tasks and Parameters Relevant for DBS As shown in Fig. 1, several tasks are chosen to assess the symptom severities. Tremor assessment includes three tasks: rest tremor, postural tremor and action tremor. Whole-hand grasping is the task used to assess bradykinesia. Passive flexion and extension of the elbow is used to assess rigidity. Each task lasts 10 s [8]. As shown in Table I, the parameters that should be extracted from sensor signal processing and displayed in the graphical user interface (GUI) are: 1) The amplitudes of tremors (R); 2) The mean range of hand grasping (peak-to-peak value of the hand grasping signals) during bradykinesia task; 3) The SD value of the hand grasping ranges during bradykinesia task; 4) Mechanical impedance (Z ), which includes viscous and elastic components, during rigidity task. All these parameters have key relations to the three primary symptoms of PD [8]. B. Static System Concept Description Fig. 2 Shows the system diagram of the glove monitoring system. A 6-axis IMU is placed on the middle finger for tremors and bradykinesia assessment. A 6-axis IMU is on the wrist and two force sensor boxes are on both sides of the wrist. These are used in the rigidity assessment. A textile glove incorporates the command module and the IMU circuit board. The command module connects with a computer via a USB cable.

Fig. 5. The flowchart of the signal processing in GMS. The glove part is based on the sensor modules and command module. The computer part is based on the program in the computer. The communication between them is the Serial-to-USB port.

A force sensor box consists of four force sensing resistors (FSRs) in parallel configuration. Fig. 3 shows the structure and prototype of a force sensor box. Four FSRs are located in the four bottom corners of the housing and are in contact with the rubber feet on its upper side. This structure has the advantage that the force sensor box gives the same value when an examiner presses on every point of the housing’s upper side with the same force. Fig. 4 shows the circuit in a force sensor box. The output of a sensor box (Vout ) is connected to an instrumentation amplifier and then connects it to the ADC input of a microcontroller. C. Solutions Process The signal processing of the glove monitoring system is described in Fig. 5. Inside the glove module, the sensor data are obtained and sent to the computer via a USB interface. Inside the computer, the received data is stored in a queue. At the same time the last cycle’s data is dequeued for analysis. Three processing

3196

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

TABLE II A LGORITHMS FOR PD S YMPTOMS S EVERITY ’ S Q UANTIFICATION

TABLE III C OEFFICIENT AND S CALING FACTORS OF THE T REMOR R EGRESSION M ODELS

methods realize the three assessment tasks correspondingly. After each assessment task, the parameters according to UPDRS are listed in the recording list. The maximum value and the minimum value of parameters are kept at the bottom of the recording list automatically. Table II shows the quantitative models of the three symptoms of PD. For the rigidity task, viscous and elastic components should be calculated ’before the calculation of mechanical impedance. IV. M ATERIALS AND M ETHODS A. Materials Two 6-axis motion tracking devices (MPU6150, InvenSense Inc.) are used as IMU sensor. It combines a 3-axis gyroscope and a 3-axis accelerometer on single silicon chip in 4 mm × 4 mm × 0.9 mm dimensions. The microcontroller in the command module reads the sensor data via an IIC interface. FSRs (FSR149, IEE Inc.) around the wrist are used to detect the forces exerted on the both sides of the wrist. The sensors consist of electrically conductive ink and a coal covering on the sheet, with a diameter of 7.5 mm. Its resistance decreases from 10 K to 1 K as the applied force goes from 1 N to 100 N. B. Processing Method of Tremor Amplitude The output of the 3-axis accelerometer can be expressed as a x yz :  (1) ax yz = ax2 + a 2y + az2 − g, where g represents gravitational acceleration and equals 9.81 m/s2 . Then there is only one axis acceleration data for the following signal processing. The dominant frequency can be calculated using PSD estimation. If the dominant frequencies in different axes are not the same, the valid dominant frequency in the axis with highest peak power is regarded as the dominant frequency of all axes. Then the total peak power in all four axes, which includes ax yz and 3-axis gyroscope signals, is the power of all axes data around the valid dominant frequency with the PSD method. The peak power in all axes after normalization is regarded as the amplitude of tremor. Heldman et al.. revealed that the logarithm of the peak power in all 3-axis accelerations and 3-axis angular velocities correlates well with the clinical scores of tremors [8].

Fig. 6. Quantification of bradykinesia during clinical hand grasp task. Here peak-to-peak angles, which calculated with a peak detection algorithm, mean the peak to peak values of all hand grasping cycles.

Because the accelerometer and gyroscope are used to measure linear and rotational movement respectively, the accelerometer realizes higher correlation for some tremor tasks, while the gyroscope performs higher correlation for other tremor tasks. Then a multiple linear regression model is used to fit the clinical ratings (UPDRS tremor scores) and the peak powers during each tremor task [22]: R = R0 + b0 · P A x yz + cx · PG x + c y · PG y + cz · PG z , (2) where R : Tremor score according to UPDRS; R0 : Coefficient of tremor score; b0 , cx , c y , and cz : Scaling factors; P A x yz : Power peak in linear acceleration; PG x , PG y , PG z : Power peak in single axis of gyroscope. R0 and the three scaling factors are different for the three tremors (rest, posture and action). These three regression coefficients were obtained according to the Stevens’ power law in psychophysics [28] Table III shows the coefficients and scaling factors for different tremor tasks. In future, the parameters in Table III will be modified according to the results of measurements in PD patients.

C. Processing Method of Bradykinesia Parameters As shown in Fig. 6, the angular displacement during hand grasping can be calculated by numerical integration of the 3-axis angular velocities at the end of middle finger. Then the mean value and SD value (|α| and σ|α| ) of hand grasping ranges can be acquired by statistical methods.

DAI et al.: NOVEL GLOVE MONITORING SYSTEM USED TO QUANTIFY NEUROLOGICAL SYMPTOMS

Fig. 7. Rigidity assessment task, where lis the length of the subject’s arm. The examiner flexes and stretch the elbow wrist through the rigidity assessment cuff attached to the wrist. Several cycles are performed during the 10 s assessment task. The examiner should make sure the elbow position of the subject is stable.

3197

Fig. 8. Prototype of the glove part. 1: command module; 2: rigidity assessment module. The command module and sensors have not direct contact with human body.

D. Processing Method of Rigidity Parameters Elastic stiffness depends on the torque and angular displacement. If a joint shows viscous behavior, it means that the measured torque depends on movement velocity. In order to avoid modeling the viscous component, some research groups chose to maintain a constant velocity by using motor actuated systems or advising examiners to impose same movement on all subjects. This GMS does not utilize motor part in order to keep the device portable and not to change the clinical assessment course. Hence the viscous and elastic components are modeled in this study [26], [29]. As shown is Fig. 7, rigidity assessment is realized by the measurement of elbow angular movement and torque on the wrist [29]. The output of two force sensors and an IMU (a 3-axis accelerometer and a 3-axis gyroscope) around the wrist are: ˙ , [F1 , F2 , a, α]

(3)

where F1 and F2 are the outputs of force sensor box 1 and force sensor box 2, respectively; and are the acceleration and angular velocity of elbow movement, respectively. Angular displacement is calculated from [a, α] ˙ by using the Direction Cosine Matrix (DCM) fusion algorithm [30]. The torque (T ) in a rigidity assessment task can be expressed as: ˙ + e, T = (F1 − F2 ) · l = c · |α| + d · |α|

(4)

where l is the subject’s arm length; c is the elastic stiffness of the subject’s elbow; d is the viscous stiffness (viscosity) of the subject’s elbow; and e is the constant offset of the sensors. Arm length (l) must be determined before the measurement, and set in the system. With 10 s duration and 100 Hz sampling rate, 1000-point data array with the format of (3) are obtained during a single rigidity assessment task. Elastic stiffness and viscosity are calculated using a leastsquares parametric method to solve (4) [29]. Mechanical impedance is the feature for UPDRS rigidity score, and is calculated as follows: Z = c + d · ω = c + d · 2π · f,

(5)

where f is the frequency of hand movement and is obtained with peak detection algorithm from the angular displacement of elbow movement.

Fig. 9. LabVIEW-based GUI of the glove measuring system. The raw data, PSD result parameters are displayed. In bradykinesia task, the angular displacement signal of hand grasping is also displayed in the bottom right corner. The results are stored in three recording lists in regular sequence.

E. System Implementation As shown in Fig. 8, at present, two prototypes are built. One is for tremors and bradykinesia assessment. Another is used for rigidity assessment. They will be combined into a single system in the next step. A 30-pin connector is used for the communication between the computer and the microcontroller [31], [32]. The GUI was realized using LabVIEW 2010 Evaluation Version. LabVIEW has the advantage of user-friendly graphical interface and of the integrated software environment. LabVIEW can be integrated with MATLAB. The DCM algorithm and (4) are realized with MATLAB programs. The raw data are stored in the computer automatically. Fig. 9 shows the software GUI of the glove monitoring system. V. VALIDATION OF A NALYTICAL M ETHODS The signals of IMU are not position information, but most research uses IMU raw data to signal processing in tremors and bradykinesia assessments. There is no previous study which has compared the assessment systems for neurological symptoms based on IMU with other medical motion tracking systems. Before comparing this system’s output with the patients’ clinical ratings, it is necessary to validate the technical values computed using the IMU tracking system which are expected to correlate to a clinician’s rating. For the rigidity assessment, the force and the elbow’s angular displacement have to be verified. In order to do this, several experiments were carried out. Analytical validations of the three test tasks of GMS were performed respectively.

3198

Fig. 10. Experimental setup for the tremor task. The test subject performed “pill-rolling” action of the hands, which means the deliberate action of rolling a pill with the thumb and index finger.

An NDI Aurora electromagnetic (EM) spatial measurement system was used as the reference system [33]. The Aurora EM system is a navigation technology designed specifically for medical applications. Aurora’s micro six degree-of-freedom (DOF) EM sensor has a small dimension (0.8 mm × L9 mm) and can be attached to a finger or wrist.

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

(a)

(b)

Fig. 11. The raw signals measured with both EM and GSM during one time imitated tremor task. (a) EM 3-axis position signals (unit: mm); (b) IMU signals: 3-axis gyroscope signal and one axis acceleration signal (gyroscope unit: °/s; accelerometer unit: g).

A. Validation of Tremor Amplitude and Frequency As shown in Fig. 10, together with the IMU, a six DOF EM sensor was attached in middle finger for real-time finger motion tracking. Because of the limited electromagnetic tracking dimension, only the postural tremor task was performed by nine subjects. Each subject performed simulative tremor actions three times with different amplitudes (slight or no tremor, moderate, and severe). The GMS and Aurora EM system were running at the same time during the experiments. And the raw sensor data from the EM sensor and IMU were stored in the PC in real-time. Fig. 11 shows the 10 s signal waveform of both the EM sensor (position) and IMU (angular velocity and acceleration). According to (1), the acceleration in three axes was combined into an absolute linear acceleration and gravity acceleration was removed. The values obtained from the prototype system (GMS) were compared with ones calculated from the EM system (positional data) with the PSD estimation method. Dominant frequency and peak power of the two systems, regarded as tremor amplitude, were compared. Because of the limited accuracy and sample rate of the EM system, neither the normal situation (no tremor or slight tremor) nor the high frequency tremors could be compared across these two systems. The dominant frequencies between the two systems have little difference in the range from 2 to 6.5 Hz, even when the subjects did not perform a movement with stable tremor frequency. For the postural tremor task, the dominant frequency of EM data is plotted against the dominant frequency of GSM in Fig. 12. The maximum, mean, and standard deviation of the frequency difference between these two systems were 0.570 Hz, 0.115 Hz, and 0.144 Hz, respectively. No linear correlation between the two systems’ peak powers could be shown. The reason for this is believed to be that seven subjects did not perform the imitated tremor activity with a consistent amplitude and frequency during each timed task.

Fig. 12. The correlation of dominant frequency in tremor assessment between EM and GMS (unit of the dominant frequency: Hz). The frequencies were calculated with PSD estimation from nine subjects.

According to the experimental results, inconsistent movement makes the power peak of the EM signals smaller when using the PSD method. When some sampling points in the EM signal are missed, the power peak increases. These two points are important for the calculation of tremor amplitude. For the two subjects who performed a stable movement, which means consistent amplitude and frequency as shown in Fig. 11, the correlation between the two power peaks of EM and GSM is shown in Fig 13. The relation between IMU signal peak power and IMU signal consistency in PD patients should be studied later. In our new program, the RMS values of all sensor data will be calculated. Together, the RMS values and peak powers can be used to determine the tremor amplitude. B. Validation of Hand Grasping Angles As shown in Fig. 6, only the gyroscope signal is used to calculate the bradykinesia parameters. Thus, errors in grasping angles are the gyroscope drift and integration error. To reduce them in a short time period, the gyroscope is calibrated upon system initiation. Three subjects performed hand grasping movements with different frequencies. Fig. 14 shows the raw data of EM and GMS for a 10 s bradykinesia task.

DAI et al.: NOVEL GLOVE MONITORING SYSTEM USED TO QUANTIFY NEUROLOGICAL SYMPTOMS

Fig. 13. The relationship of the tremor amplitudes (power peak) acquired with both EM and GMS. Two subjects each performed four imitated tremor tasks. The imitated tremor amplitudes varied in range from slight to severe for each subject. The power peak of GMS calculated from both gyroscope and accelerometer signals with the PSD method, while the power peak of EM was calculated only from the position signal of the subject’s finger. This plot shows that the correlation between these two systems corresponds approximately to linear.

Fig. 14. Raw sensor data both in EM (location and orientation data) and GSM (angular velocity) during one time bradykinesia task of a subject.

TABLE IV A BSOLUTE D ISAGREEMENT B ETWEEN EM T RACKING AND

3199

Fig. 15. Experimental setup for rigidity task. The subject performed the same elbow movement as during the rigidity task. Two EM sensors were attached to the wrist.

Fig. 16. Plot of DCM fusion (IMU tracking) versus EM tracking during passive elbow movement. The elbow movement during a rigidity assessment task was measured with both IMU and EM tracking. IMU tracking values were realized with the DCM algorithm from gyroscope and accelerometer signals. EM tracking values were obtained directly from the EM sensors.

algorithm can be used in calculating the hand grasping angle range.

IMU T RACKING (B RADYKINESIA TASK )

C. Validation of Wrist Angles During Rigidity Assessment

Then the EM signals (three-dimensional orientation) were processed with the fast Fourier transform (FFT) method to get frequency and the peak-detection method to obtain the grasping range. Table IV shows the disagreement between the EM system and the IMU-based glove monitoring system. Because there were some bad fits (inaccurate) in the results of the EM system during hand grasping movements, another tracking system, such as an optical tracking system, can be used as the reference system in the next step. In order to improve the accuracy of bradykinesia parameters, further signal-processing methods must be carried out. The integration method of angular velocity plays a key role in the calculation of bradykinesia parameters. Also a DCM

As shown in Fig. 15, two six DOF EM sensors attached to the wrist were used in the elbow-motion-tracking experiment. With the help of an examiner, the subject performed elbow flexion and extension movements around the z axis, and the end of the elbow worked as a fulcrum. With the position and orientation data from two EM sensor points, the elbow’s angular movement was calculated and compared with the elbow angle position calculated with the DCM method using the IMU data. Six subjects performed the rigidity task during the experiment. In one subject, too many sampling points from the EM data were missing; thus only data from five subjects was used. Fig. 16 shows the elbow’s movement angles both from the DCM fusion algorithm (IMU tracking) and from EM tracking in the second subject. Their difference and maximum difference range over 10 s are also shown in Fig. 16. Table V shows the absolute disagreement between the EM system and the IMU system. The maximum difference, mean difference and standard deviation of the differences in all data are 12.50°, -4.74°, and 1.91°, respectively.

3200

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

TABLE VI PARAMETERS FOR THE T WO -T ERM G AUSSIAN R EGRESSION

Fig. 17. Force sensor box test with weight. The weight was placed in the center top side of the force sensor box.

E. Validation of Mechanical Impedance Components TABLE V A BSOLUTE D ISAGREEMENT B ETWEEN EM T RACKING AND IMU T RACKING

D. Validation of Force Sensor Boxes According to Fig. 4, the force-to-voltage conversion is tied to a measuring resistor in a voltage divider configuration. The output voltage of the force sensor box (Vout ) is shown as in Vout = V+ / [1 + (R F S R /R M )] ,

(6)

where V+ is the power supply (5 V); R F S R is the resistor value of the force sensor box and R M is the pull down resistor (4.7 K). Equation (6) shows that the output of the force sensor box response for R F S R is similar to an e-function. The forceresistance characteristic of a single FSR (FSR149) is nonlinear and the response approximately follows an inverse powerlaw characteristic. But a linear force-to-voltage relationship is needed in (4). Hence, a regression analysis was performed to get the voltage-force function. As shown in Fig. 17, 58 weight loads ranging from 10 grams to 9000 grams were exerted on the center of the top side of each force sensor box. Then 116 voltage values were acquired. Using the Matlab curve fitting toolbox, 1-, 2- and 4-term Gaussian, e-Function, and Polynomial regression analyses were performed with the experimental data. A 2-term Gaussian function, as shown in (7), has the highest correlation with the experimental data: F(vout ) = a1 · e− (

vout − b1 2 vout − b2 2 ) + a 2 · e− ( ) , c1 c2

(7)

where F is the calculated force value; Vout is the sensor box output voltage; a, b, and c are coefficients. The coefficients of (7) can also be determined using the MATLAB curve fitting toolbox. Table VI shows the coefficients for both force sensor boxes. Another experiment was carried out to get the deviation between F and its real value (the value of weight). When the weights, ranging from 2.47 to 30.05 N, were applied to the center of both sensor boxes ten times, the average deviations of two force sensor boxes were 6% and 8%, respectively.

In order to acquire the relation between mechanical impedance and UPDRS scores, first the relation of elastic stiffness and viscosity with the rigidity severity should be investigated. Nine healthy volunteers (average age: 24.4 ± 4.2 years) were tested with the system, yielding eight measurements each. During the first four movements, volunteers were asked to relax (with no rigidity and as the reference); while in the next four movements they were asked to imitate rigidity. The viscosity values, elasticity values, and frequencies during the measurements were calculated with the algorithms stated above. The viscosity and elasticity were converted to absolute values before calculation. The average value of viscous modulus with no rigidity (relaxed state) was 0.26 ± 0.08 N·m/degree, and 0.78 ± 0.45 N·m/degree in the imitated rigidity state. The mean value of elastic modulus with no rigidity (relaxed state) was 0.99 ± 0.53 N·m/degree, and 3.78 ± 2.85 N·m/degree in the imitated rigidity state. For Cohen’s d, an effect size of 0.8 to infinity denotes a “large” effect [34]. The effect size (Cohen’s d) of viscosity and elasticity between normal state and imitated rigidity were, therefore, “large”, 1.61 and 1.36, respectively.

VI. C ONCLUSION The concept of a portable glove monitoring system used in DBS and two prototypes were presented. This system can assess all the three primary PD symptoms and list them in a GUI. Thus, the surgeon is supported in finding the best position for the stimulating electrode. All the sensors have no contact with the human body. In comparative experiments with an EM tracking system, we showed that the IMU and force sensors work well. Nevertheless, attention must be paid to the signal processing methods. 1) The influence of signal noise in tremor assessment when the PD patient has a slight tremor. 2) Inconsistent movement and its influence in tremor amplitude during the tremor assessment task, especially for peak power calculation with PSD estimation. 3) The combination of six axes signals in both gyroscope and accelerometer, which represent the rotational and the linear movement respectively. 4) The correlation between different sensor signals and the UPDRS ratings. 5) The repeatability of parameters with the same patient at different times.

DAI et al.: NOVEL GLOVE MONITORING SYSTEM USED TO QUANTIFY NEUROLOGICAL SYMPTOMS

A better cognition of the analytical methods in this system was achieved with these validation results, providing the necessary science and engineering to guide future signal processing methods. Then the clinical experiments can be performed. The next steps also consist in carrying out measurements with PD patients in order to find out the scale factors and coefficients in (2) correlating the system output with the clinician’s rating. For bradykinesia and rigidity tasks, the correlation between assessed parameters using sensors and the UPDRS scores are needed to be further investigated. Afterwards, the system output must be validated with another set of patients evaluated by a clinician. In addition, a new force sensor, which has better performance than FSRs, can be adopted in this system [36]. For the assessment of DBS side effects such as dyskinesia, paresthesia, and other non-motor symptoms, more parameters should be realized in the next version [37]. ACKNOWLEDGMENT The authors would like to thank Prof. Dr. Tonn from the Department of Neurosurgery, University Hospital Grosshadern, (LMU München) for their support in this project. R EFERENCES [1] M. S. Okun and K. D. Foote, “Parkinson’s disease DBS: What, when, who and why? The time has come to tailor DBS targets,” Expert Rev. Neurotherapeutics, vol. 10, no. 12, pp. 1847–1857, Dec. 2010. [2] P. Crawford and E. E. Zimmerman, “Differentiation and diagnosis of tremor,” Amer. Family Phys., vol. 83, no. 6, pp. 697–702, Mar. 2011. [3] G. Deuschl, P. Bain, and M. Brin, “Consensus statement of the movement disorder society on tremor,” Movement Disorders, vol. 13, no. 3, pp. 2–23, 1998. [4] R. Phwa, K. E. Lyons, and W. Koller, Handbook of Parkinson’s Disease, 4th ed. The Switzerland: Informa Healthcare, 2007, pp. 53–59. [5] R. G. Bittar, S. C. Burn, P. G. Bain, S. L. Owen, C. Joint, D. Shlugman, and T. Z. Aziz, “Deep brain stimulation for movement disorders and pain,” J. Clinical Neurosci., vol. 12, no. 4, pp. 457–463, May 2005. [6] S. M. Rissanen, M. Kankaanpaa, M. P. Tarvainen, and V. Novak, “Analysis of EMG and acceleration signals for quantifying the effects of deep brain stimulation in Parkinson’s disease,” IEEE Trans. Biomed. Eng., vol. 58, no. 9, pp. 2545–2553, Jun. 2011. [7] S. Rissanen1, M. Kankaanpää, M. P. Tarvainen, J. Nuutinen, I. M. Tarkka, O. Airaksinen, and P. A. Karjalainen “Analysis of surface EMG signal morphology in Parkinson’s disease,” Phys. Meas., vol. 28, no. 12, pp. 1507–1521, Dec. 2007. [8] J. P. Giuffrida, D. E. Riley, B. N. Maddux, and D. A. Heldman, “Clinically deployable Kinesia technology for automated tremor assessment,” Movement Disorders, vol. 24, no. 5, pp. 723–730, Apr. 2009. [9] A. Salarian, H. Russmann, C. Wider, P. R. Burkhard, F. J. G. Vingerhoets, and K. Aminian, “Quantification of tremor and bradykinesia in Parkonson’s disease using a novel ambulatory monitoring system,” IEEE Trans. Biomed. Eng., vol. 54, no. 2, pp. 313–322, Feb. 2007. [10] K. Niazmand, K. Tonn, A. Kalaras, U. M. Fietzek, J. H. Mehrkens, and T. C. Lueth, “Quantitative evaluation of Parkinson’s disease using sensor based smart glove,” in Proc. IEEE Symp. CBMS, Jun. 2011, pp. 1–8. [11] K. Niazmand, C. Jehle, L. T. D’Angelo, and T. C. Lueth, “A new washable low-cost garment for everyday fall detection,” in Proc. IEEE Annu. Int. Conf. EMBS, Sep. 2010, pp. 6377–6380. [12] A. Czabke, L. T. D’Angelo, K. Niazmand, and T. C. Lueth, “Ein kompaktes system zur erfassung und dokumentation von bewegungs gewhohheiten,” in Proc. Conf. AAL, 2009, pp. 1–5. [13] C. N. Riviere, S. G. Reich, and N. V. Thakor, “Adaptive fourier modeling for quantification of tremor,” J. Neurosci. Methods, vol.74, no. 1, pp. 77–87, Jun. 1997.

3201

[14] Z. Yuting, S. Markovic I. Sapir, R. C. Wagenaar, and T. D. C. Little, “Continuous functional activity monitoring based on wearable tri-axial accelerometer and gyroscope,” in Proc. IEEE Pervas. Health Conf., vol. 99. May 2011, pp. 1–9. [15] G. Rigas, A. T. Tzallas, D. G. Tsalikakis, S. Konitsiotis, and D. I. Fotiadis, “Real-time quantification of resting tremor in the Parkinson’s disease,” in Proc. IEEE Annu. Int. Conf. EMBS, Sep. 2009, pp. 1306–1309. [16] B. Wingeier, T. Tcheng, M. M. Koop, B. C. Hill, G. Heit, and H. M. Bronte-Stewart, “Intra-operative STN DBS attenuates the prominent beta rhythm in the STN in Parkinson’s disease,” Exp. Neurol., vol. 197, no. 1, pp. 244–251, Nov. 2006. [17] B. K. Park, J. W. Kim, Y. Kwon, J.-W. Kim, and J.-H. Lee, “Analysis of viscoelastic properties of wrist joint for quantification of parkinsonian rigidity,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 19, no. 2, pp. 167–176, Apr. 2011. [18] S. J. M. Bamberg, A. Y. Benbasat, D. M. Scarborough, D. E. Krebs, and J. A. Paradiso, “Gait analysis using a shoe-integrated wireless sensor system,” IEEE Trans. Inf. Technol. Biomed., vol. 12, no. 4, pp. 413–423, Jul. 2008. [19] J. W. Kim, J. H. Lee, Y. Kwon, C.-S. Kim, G.-M. Eom, S.-B. Koh, D.-Y. Kwon, and K.-W. Park, “Quantification of bradykinesia during clinical finger taps using a gyrosensor in patients with Parkinson’s disease,” Med. Biol. Eng. Comput., vol. 49, no. 3, pp. 365–371, Mar. 2011. [20] L. T. D’Angelo, A. Czabke, K. Niazmand, and T. C. Lueth, “ART-a new concept for an activity recorder and transceiver,” in Proc. IEEE EMBS Conf., Jan. 2010, pp. 2132–2135. [21] T. C. Lueth, L. T. D’Angelo, and A. Czabke, “Chapter. 14: TUM-AgeTech: A new framework for pervasive medical devices,” in Pervasive and Smart Technologies for Healthcare: Ubiquitous Methodologies and Tools, New York, NY, USA: IGI Global, 2010, pp. 295–321. [22] R. J. Elble, S. L. Pullman, J. Y. Matsumoto, J. Raethjen, G. Deuschl, and R. Tintner, “Tremor amplitude is logarithmically related to 4- and 5-point tremor rating scales,” Brain, vol. 129, no. 10, pp. 2660–2666, Jul. 2006. [23] G. Mostile, J. P. Giuffrida, O. R. Adam, A. Davidson, and J. Jankovic, “Correlation between Kinesia system assessments and clinical tremor scores in patients with essential tremor movement,” Movement Disorders, vol. 25, no. 12, pp. 1938–1943, Sep. 2010. [24] D. A. Heldman, J. Jankovic, D. E. Vaillancourt, J. Prodoehl, R. J. Elble, and J. P. Giuffrida, “Essential tremor quantification during activities of daily living, “ Parkinsonism Rel. Disorders, vol. 17, no. 7, pp. 537–542, Aug. 2011. [25] B. Post, M. P. Maruschka, R. M. Bie de, R. J. de Haan, and J. D. Speelman, “Unified Parkinson’s disease rating scale motor examination: Are ratings of nurses, residents in neurology, and movement disorders specialists interchangeable?,” Movement Disorders, vol. 20, no. 12, pp. 1577–1584, Dec. 2005. [26] S. M. Charles, Parkinson’s Disease: Overview and Current Abstracts, 1st ed., New York, NY, USA: Nova Sci., 2003. [27] D. A. Heldman, J. P. Giuffrida, R. Chen, M. Payne, F. Mazzella, A. P. Duker, A. Sahay, S. J. Kim, F. J. Revilla, and A. J. Espay, “The modified bradykinesia rating scale for Parkinson’s disease: Reliability and comparison with kinematic measures,” Movement Disorders, vol. 26, no. 10, pp. 1859–1863, Aug. 2011. [28] W. Ellermeier and G. Faulhammer, “Empirical evaluation of axioms fundamental to Stevens’s ratio-scaling approach: I. Loudness production,” Perception Psychophys., vol. 62, no. 8, pp. 1505–1511, Dec. 2000. [29] S. K. Patrick, A. A. Denington, M. J. Gauthier, D. M. Gillard, and A. Prochazka, “Quantification of the UPDRS rigidity scale,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 9, no. 1, pp. 31–41, Mar. 2001. [30] S. O. H. Madgwick, A. J. L. Harrison, and A. Vaidyanathan, “Estimation of IMU and MARG orientation using a gradient descent algorithm,” in Proc. IEEE ICORR Conf., Jun. 2011, pp. 1–7. [31] L. T. D’Angelo, S. Michael, N. Paul, and T. C. Lueth, “A sensor network to iPhone interface separating continuous and sporadic processes in mobile telemedicine,” in Proc. IEEE EMBS Conf., Sep. 2011, pp. 1528–1531. [32] Y. Su, C. R. Allen, D. Geng, and D. Burn, “3-D motion system (‘datagloves’): Application for Parkinson’s disease,” IEEE Trans. Instrum. Meas., vol. 52, no. 3, pp. 662–674, Jun. 2003. [33] (Aug. 2012). NDI Aurora Electromagnetic Tracking System [Online]. Available: http://www.ndigital.com/aurora.php

3202

IEEE SENSORS JOURNAL, VOL. 13, NO. 9, SEPTEMBER 2013

[34] S. Nakagawa and I. C. Cuthill, “Effect size, confidence interval and statistical significance: A practical guide for biologists,” Biol. Rev. Cambridge Phil. Soc., vol. 82, no. 4, pp. 591–605, Nov. 2007. [35] L. J. Bour, M. F. Contarino, E. M. Foncke, R. M. de Bie, P. van den Munckhof, J. D. Speelman, and P. R. Schuurman, “Long-term experience with intraoperative microrecording during DBS neurosurgery in STN and GPi,” Acta Neurochirurgica, vol. 152, no. 12, pp. 2069–2077, Dec. 2010. [36] A. Massaro, F. Spano, A. Lay-Ekuakille, P. Cazzato, R. Cingolani, and A. Athanassiou, “Design and characterization of nanocomposite pressure sensor implemented in tactile robotic system,” IEEE Trans. Instrum. Meas., vol. 60, no. 8, pp. 2967–2975, Aug. 2011. [37] A. Lay-Ekuakille, P. Vergallo, A. Trabacca, M. De Rinaldis, F. Angelillo, F. Conversano, and S. Casciaro, “Low-frequency detection in ECG signals and joint EEG-ergospirometric measurements for precautionary diagnosis,” Measurement, vol. 46, no. 1, pp. 97–107, Jan. 2013.

Houde Dai received the master’s degree in biomedical engineering from Tongji University, Shanghai, China, in 2009. He is currently pursuing the Ph.D. degree at the Institute of Micro Technology and Medical Device Technology, TU Muenchen, Freising, Germany. His work includes quantitative assessment of neurological symptoms based on inertial sensors and force sensors. His current research interests include medical electronic device development and biomedical signal processing.

Bernward Otten is currently pursuing the master’s degree in mechanical engineering with the Faculty of Mechanical Engineering, TU Muenchen, Freising, Germany. He received the B.Sc. degree in medical technology from TU Muenchen in 2012. His current research interests include control theory.

Jan Hinnerk Mehrkens received the M.D. degree in radiology from the University of Heidelberg, Heidelberg, Germany, in 2000. He is the Leader of the Functional Neurosurgery Group, Neurosurgical Clinic, University of Munich, Munich, Germany. His current research interests include functional neurosurgery for uncontrollable pain syndromes, movement disorders, and spasticity. He has received speaker honoraria from Medtronic Inc.

Lorenzo T. D’Angelo received the M.D. degree in radiology from the University of Heidelberg in 2000. His research interests cover functional neurosurgery for uncontrollable pain syndromes, movement disorders, and spasticity. He has received speaker honoraria from Medtronic Inc. He received the Dipl.Ing. (M.Sc.) degree in mechanical engineering from TH Karlsruhe, Karlsruhe, Germany, in 2006, and the Ph.D. degree in mechanical engineering from TU Muenchen, Freising, Germany, in 2011. Currently, he leads the AgeTech Group, developing personal assistive devices for elderlies, the Department of Micro Technology and Medical Device Technology, TU Muenchen. His current research interests include the development of event-based reminding systems, mobility aids and exoskeletons for locomotion, transfer and manipulation support as well as systems for automatic fluid intake assessment, and dehydration prevention.

Tim C. Lueth (M’89) received the Dipl.-Ing. degree in electrical engineering from the Darmstadt University of Technology, Darmstadt, Germany, in 1989, and the Ph.D. degree in robotics and habilitation in computer science from the University of Karlsruhe, Karlsruhe, Germany, in 1993 and 1997, respectively. From 1994 to 1995, he was a Visiting Researcher with the MITI-AIST Electrotechnical Laboratory, Tsukuba, Japan. In 1997, he became a Professor of surgical navigation and robotics with the Medical School Charité-Universitätsmedizin Berlin, Humboldt University, Berlin, Germany. In 2001, he became the Director for Mechatronic Medical Technology at the Fraunhofer-Institute for Production Systems and Design Technology, Berlin. Since 2005, he has been a Professor, Chair, and Director with the Institute of Micro Technology and Medical Device Technology, University of Technology, Munich, Germany. In 2006, he was a Professor with the University of Toronto, Toronto, ON, Canada. He elected a TOP-3 Inventor from the European Patent Office in the category “lifetime achievement” for his patent activities in the area of surgical robotics and navigation, in 2007. He received several national and international awards for his research on medical devices. In 2010, he became an Elected Member of “acatech,” from the German National Academy for Science and Technology. He is an active member of the IEEE R&A Chapter and the IEEE Engineering in Medicine and Biology Society Chapter.