Computerized Measurement of Motor Performance After Stroke
Background and Purpose Stroke scales usually convert motor status to a score along an ordinal scale and do not provide a permanent recording of motor performance. Computerized methods sensitive to small changes in neurological status may be of value for studying and measuring stroke recovery.
Methods We developed a computerized dynamometer and tested 23 stroke subjects and 12 elderly control subjects on three motor tasks: sustained squeezing, repetitive squeezing, and index finger tapping. For each subject, scores on the Fugl-Meyer and National Institutes of Health stroke scales were also obtained.
Results Sustained squeezing by the paretic hand of stroke subjects was weaker (9.2 kg) than the unaffected hand (20.2 kg; P<.0005), as well as control dominant (23.1 kg; P<.0005) and nondominant (19.9 kg; P<.005) hands. Paretic index finger tapping was slower (2.5 Hz) than the unaffected hand (4.2 Hz; P<.01), as well as control dominant (4.7 Hz; P<.0005) and nondominant (4.9 Hz; P<.0005) hands. Many features of dynamometer data correlated significantly with stroke subjects’ Fugl-Meyer scores, including sustained squeeze maximum force (ρ=.91) and integral of force over 5 seconds (ρ=.91); repetitive squeeze mean force (ρ=.92) and mean frequency (ρ=.73); and index finger tap mean frequency (ρ=.83). Correlation of these motor parameters with National Institutes of Health stroke scale score was weaker in all cases, a consequence of the scoring of nonmotor deficits on this scale. Dynamometer measurements showed excellent interrater (r=.99) and intrarater (r=.97) reliability.
Conclusions The degree of motor deficit quantitated with the dynamometer is strongly associated with the extent of neurological abnormality measured with the use of two standardized stroke scales. The computerized dynamometer rapidly measures motor function along a continuous, linear scale and produces a permanent recording of hand motor performance accessible for subsequent analyses.
Hemiparesis constitutes the most common neurological abnormality after stroke, with a wide range in the severity of the motor deficit.1 A number of standardized scales have been devised to measure stroke deficits. These scales have widespread utility, including monitoring recovery, assessing therapeutic interventions, and prognostication.2 Each scale has its limitations, however.3–5 Some are insensitive to small deficits or require special skills to administer. The motor performance itself is not actually recorded but is instead converted to a discrete score along an ordinal scale. Most scales thus suffer from a lack of linearity.5
A device capable of rapidly measuring motor output along a linear scale may be of value in the evaluation of stroke recovery. The ability to subsequently review the performance or perform computational analyses would also be useful for obtaining a more detailed understanding of motor status after stroke. Therefore, we developed a computerized dynamometer to digitalize, store, and analyze motor performances. Previous studies of stroke patients have reported that maximum tapping frequency6,7 and squeezing strength7–9 are reduced in paretic hands compared with either unaffected hands or control hands. However, quantitative motor analysis has not previously been compared directly with other neurological assessments. The goals of the present study were to assess the extent to which dynamometer measurements of squeezing and finger tapping (1) distinguish the paretic hand of stroke subjects from the unaffected hand of stroke subjects and control hands and (2) correlate with scores on reliable10,11 neurological scales.
Subjects and Methods
The main components of the dynamometer are a hand-held force transducer, an amplifier, an analog-to-digital converter, and a computer with strip chart–emulating software (Fig 1⇓). Two force transducers (SSM-AJ-50 and SSM-AJ-250, Interface) were used: one with a 23-kg (50-lb) limit, used for tapping studies, and one with a 114-kg (250-lb) limit, used for squeezing studies. Each transducer was fitted with smooth, round, plastic handles. Transducers were 9×8×3 cm, with a grip circumference of 18 cm and a weight of 180 g. Each transducer converted isometrically generated force into an electric signal. This signal was carried to an amplifier (ETH-200, AD Instruments), then to an analog-to-digital converter (MacLab, AD Instruments). The digital output was fed into a Macintosh computer (PowerBook 540c or 5300cs, Apple Computer) running strip chart–emulating software (Chart v. 3.5.1, AD Instruments).
The computer, amplifier, and MacLab all plugged into a single extension outlet. This outlet was serially connected to an isolation transformer followed by a standard wall outlet, the transformer having been inserted to prevent electrical injury to test subjects. All of the apparatus was housed inside the shell of an electrocardiogram cart, allowing for transportation to the bedside.
Stroke subjects were identified over a 10-month period from the admissions records of Massachusetts General Hospital and Spaulding Rehabilitation Hospital, as well as from outpatient departments. Elderly control subjects were recruited through local advertisements. The entry criterion for control subjects was the absence of active neurological disease. The entry criterion for stroke subjects was a history of a stroke producing any hand weakness. Exclusion criteria, defined by National Institutes of Health (NIH) stroke scale questions, were decreased level of consciousness (any points on questions 1a, 1b, or 1c), aphasia (≥2 points on question 9), or neglect (2 points on question 11). Approximately 1 in 3 stroke patients was eligible for this study according to these criteria. Subjects provided informed consent and then received dynamometer testing plus measurement of scores on two reliable10,11 stroke scales. One was the NIH stroke scale, a global scale designed to measure deficits in a number of neurological domains. A form modified to be sensitive to distal extremity weakness was used12; possible scores range from 0 (normal) to 44. The second scale was the Fugl-Meyer (FM) arm motor subscore,13 a motor scale that assesses multiple features of upper extremity motor function. Possible scores on this scale range from 0 to 66 (normal). All dynamometer measurements and clinical scale assessments were performed by two of the authors (S.C.C. and G.N.) after being instructed by NIH and FM scale videos. Handedness was evaluated with the use of the Edinburgh Inventory.14
Settings used for the strip chart–emulating software recording were acquisition of 40 to 100 data points per second, a low-pass filter of 50 Hz, and a high-pass filter of .05 Hz. Before each dynamometer use, the amplifier readout was set to zero with the transducers flat on a horizontal surface. The voltage output of the dynamometer was converted to kilograms by measuring the millivolt deflection occurring with suspension of two 100-g weights (McMaster-Carr) from the ends of the dynamometer. The linearity of the conversion was then confirmed by suspending two 2-kg weights. Subsequently, these weights were used to assess the fidelity of calibration every other week, with no deviation identified during the course of the study.
Dynamometer examination methods were based on previous recommendations.15 The subject was positioned either sitting or lying in bed with head elevated to 45°. Three tasks were performed with each hand: (1) sustained squeezing, (2) repetitive squeezing, and (3) index finger tapping. For squeezing, arms were placed onto a table with elbows flexed to 90° and wrists in neutral position. For tapping, the same position was used for the first 9 stroke subjects. Under these conditions, however, a variable number of upper extremity joints were used to tap. A support was constructed onto which the subject’s forearm was secured by means of mid-ulna and mid-metacarpal hook and loop fastener straps. This restricted tapping movements to the index finger metacarpophalangeal joint and was used in subsequent subjects.
Squeezing was examined before tapping, and right hand was examined before left for each task. Subjects were given standard instructions to perform each of the following tasks until told to stop: (1) squeeze as hard as possible, (2) squeeze repetitively as fast and hard as possible, or (3) tap as rapidly as possible using only the index finger. Each subject was given an opportunity to practice, with feedback from the examiner when a task was performed improperly. Each task was then performed and measured for 7 to 10 seconds.
The examiner indicated the beginning and end of task performance verbally, simultaneously stamping these comments onto the recording using the Macintosh keyboard with a text feature of the strip chart–emulating software. Acquisition of dynamometer data for the three tasks required less than 6 minutes in most cases.
Analysis of Dynamometer Recordings
For tapping and repetitive squeezing, each recording was converted to a series of cycles. Peak cycle values and time between cycles were then exported to a statistics software program (JMP-In 3.1.5, SAS Institute). Measurements were made of mean force and mean frequency for repetitive squeezing; mean frequency for index finger tapping; and their respective coefficients of variation. For sustained squeezing, the strip chart–emulating software was used to directly measure parameters.
Dynamometer values for stroke subjects’ paretic hand motor performances were compared with unaffected hand performances as well as both hands of control subjects with the use of Student’s t test. No correction was made for multiple comparisons. Two representative motor parameters were evaluated: the maximum force during sustained squeeze and the mean frequency during index finger tap. Analysis of results on index finger tapping was restricted to the 14 stroke subjects examined while using the arm support. A normalized value was also compared between stroke and control subjects. This normalized value was generated to account for intersubject variability in squeezing and tapping.16,17 For control subjects, the normalized value was nondominant/dominant hand dynamometer values. For stroke subjects, the normalized value was paretic/unaffected hand dynamometer values.
For stroke subjects, linear correlation analyses were done to assess the degree of correlation of each dynamometer measurement with (1) the score on the FM scale and (2) the score on the NIH stroke scale. The Spearman rank-order correlation statistic, ρ, was used.
This study was approved by the Massachusetts General Hospital and the Spaulding Rehabilitation Hospital Human Studies Committees. The dynamometer was approved by the Massachusetts General Hospital Bioengineering Department.
Intrarater and interrater reliability were assessed in a separate group of 10 right-handed subjects (5 control subjects and 5 stroke subjects), each of whom met entry criteria. Three independent dynamometer studies were obtained on each subject during a 30- to 60-minute period: examiner 1 (S.C.C.) measured sustained squeezing and index finger tapping by the paretic hand (stroke subjects) or dominant hand (control subjects). These measurements were repeated by examiner 2 (J.D.S.). A final set of measurements was then obtained by examiner 1. Both examiners subsequently measured the maximum force during sustained squeezing and the mean frequency during index finger tapping from all three data sets. Reliability was assessed with the Pearson product moment correlation coefficient.
Dynamometer measurements and scores from neurological scales were obtained in 23 stroke subjects and 12 control subjects. The mean age of stroke subjects (66 years; range, 36 to 89 years) and control subjects (67 years; range, 61 to 76 years) was not significantly different (P>.5). All control subjects had normal NIH (0) and FM (66) scores. For stroke subjects, mean score on the FM scale (45) was significantly lower (P<.0005) and on the NIH scale (4) significantly higher (P<.0001) than that of control subjects. All control and 22 of 23 stroke subjects (premorbid status) were strongly right-handed; one stroke subject was ambidextrous and was treated as right-handed during analysis. There was a trend toward more men in the stroke group (18 of 23 male for stroke subjects, 5 of 12 male for control subjects; P=.059 by Fisher’s exact test).
The median time after stroke was 22 days (range, 1 to 360 days). The location of the stroke was right hemisphere in 7, left hemisphere in 10, and infratentorial in 6 cases. The paretic hand was the right in 14 cases and left in 9. The index stroke was the first cerebral infarct in 19 of 23 subjects and the first symptomatic infarct in 21 of 23. Stroke mechanism was atheroembolic in 8 cases, small vessel in 14 cases, and hemorrhagic in 1 case. All subjects were able to generate a measurable performance for each task, except for 4 subjects who could not tap the paretic index finger; the mean FM score for these 4 subjects was 27 of 66.
Comparison of Stroke Subjects and Control Subjects
Examples of dynamometer recordings for sustained squeezing and index finger tapping are shown in Fig 2⇓. Values for paretic hand were significantly lower than values for unaffected hand, control dominant hand, and control nondominant hand during both sustained squeezing maximum force (Fig 3A⇓) and index finger tapping mean frequency (Fig 3B⇓). This remained true when the side of paresis was considered during comparison with control subjects: dynamometer values for subjects with dominant hand paresis were less than control dominant hand values, and values for subjects with nondominant hand paresis were less than control nondominant hand values. The normalized values were significantly different between stroke (paretic/unaffected) and control (nondominant/dominant) subjects for both squeezing (0.50 versus 0.85; P<.005) and tapping (0.62 versus 1.04; P<.005). No significant differences were found between the unaffected hand of stroke subjects and either hand of control subjects.
Correlation of Dynamometer Values With Scores on Neurological Scales
The Table⇓ shows the results of correlating normalized values for motor parameters with scores on FM and NIH stroke scales. In all cases, use of the normalized parameter showed improved correlation compared with use of the paretic hand value alone. This is demonstrated by comparing the first two rows of the Table⇓ for the correlation between FM score and sustained squeeze maximum force: the correlation using the ratio of paretic to unaffected hand (ρ=.91) is stronger than the correlation using the paretic hand (ρ=.81). None of the coefficient of variation values, reflecting inconsistent motor effort, showed significant correlation with scores on either neurological scale. Eight of the 12 examined motor parameters showed significant correlation with the FM arm motor score. For six of these parameters, correlation with the NIH score was also significant, although at a lower level than with the FM score. Linear regressions for sustained squeeze maximum force and index finger tap mean frequency are shown in Fig 4⇓.
A high degree of intrarater reliability was found on comparing the two sets of squeezing and tapping results obtained by examiner 1 (r=.97, P<.0001). The means were 18.77 kg versus 18.01 kg for maximum squeezing force and 4.01 Hz versus 4.33 Hz for tapping frequency. Comparing squeezing and tapping results between examiners 1 and 2 identified excellent interrater reliability (r=.99, P<.0001); the means for examiner 2 were 16.67 kg for squeezing and 4.09 Hz for tapping. Both examiners subsequently extracted dynamometer parameters from all three sets of data; excellent interrater reliability for the data analysis methods was found (r=.99, P<.0001).
Two lines of evidence suggest that dynamometer measurements are a valid assessment of motor status after stroke. First, dynamometer measurements of paretic hand squeezing and tapping were lower than those obtained from the unaffected hand of stroke subjects, the dominant hand of control subjects, or the nondominant hand of control subjects (Fig 3⇑). This finding is consistent with previous reports.6–9 Second, dynamometer-derived measures of motor status correlate closely with scores on two standardized stroke scales, the FM and NIH stroke scales (Table⇑, Fig 4⇑). This relationship has not previously been described. Desrosiers et al18 examined grip strength in the unaffected hand and found no significant correlation with paretic arm FM score; correlation of affected hand strength with FM score was not presented.
For the measurement of motor status, use of this newly developed dynamometer offers several theoretical advantages over standardized clinical scales because of its use of a continuous, high-resolution, linear scale. First, although values from clinical scales are often treated as continuous variables in statistical testing,1 they are actually the sum of several ordinal variables, a comparatively weaker form of measurement.19 Second, direct measurement of squeezing and tapping produces data of higher resolution than most clinical scales. For example, three stroke subjects had an NIH score of 3. However, these three were not equivalent when tested with the dynamometer, with a range for index finger tapping frequency of 1.5 to 5.6 Hz and a range for sustained squeeze maximum force of 14.7 to 27.7 kg. The higher resolution of dynamometer testing may allow for improved ability to detect smaller changes in motor recovery over shorter time intervals. Third, there are fewer assumptions about data points in a linear scale. For example, improvement in squeezing from 20 to 30 kg represents a 50% increase in strength. A change in FM score from 20 to 30 corresponds to an undefined degree of neurological improvement.5 This advantage of a linear measurement will be valuable in future studies of stroke recovery employing serial dynamometer measurements; in this regard, the excellent intrarater and interrater reliability demonstrated for the dynamometer will also be important. Finally, a linear scale may better represent changes in motor cortex function after stroke, since the discharge rate of many cortical neurons has been shown to be linearly related to upper extremity force generation.20,21
Computational analysis of the digitalized dynamometer recordings may also permit a more detailed understanding of motor behavior than is available with the use of either standardized clinical measures or a mechanical dynamometer. This is because motor performances are recorded, permitting subsequent review and analysis. As an example, measurement of multiple parameters for squeezing (Table⇑) provides a broader picture of motor dysfunction than measurement of maximum force alone, the usual output of a noncomputerized dynamometer. The correlation of maximum force with stroke scale scores remained significant (Table⇑) whether we compared a sustained or a repetitively produced squeeze. The maximum force and the integral of force over time were affected to a similar extent after stroke. The decrement in motor output over 5 seconds (line 4 of the Table⇑) showed significant correlation with the FM score. The time to reach maximum force (Table⇑, line 5) was used to test the hypothesis that a weaker subject reaches maximum force more slowly, but this variable did not show a significant relationship with FM score. Overall, subjects with a greater motor deficit show squeezing that is weaker but not slower in its onset. If the squeeze is sustained, there will be a greater decrement of force and a smaller integral of force over time. If a repetitive volley of squeezes follows, these will be produced more slowly.
A number of variables complicate direct comparison of the current data with prior reports. Differences in the instrument used for measurement can influence motor findings.22,23 Previous studies have assessed nonisometric forces because they employed a mechanical dynamometer7–9,17 for squeezing and either a key press6 or tally counter7,17 for tapping. The current device measures squeezing and tapping isometrically. A large number of variables can influence motor performance after stroke, including side of paresis, time after stroke, lesion volume, and stroke subtype.18,24,25 Despite these variables, results of the current study were similar to prior studies. For example, Colebatch and Gandevia8 studied 10 subjects, most of whom had cerebral infarcts. Mean grip strength was 9.3 kg in the paretic hand and 27 kg in the unaffected hand, with a normalized ratio of 0.34. This is similar to the values found in the present study (Fig 3A⇑): 9.2 kg in the paretic hand, 20.2 in the unaffected hand, with a normalized ratio of 0.46. Shimoyama et al6 studied the paretic hand of 14 subjects with cerebral lesions, 12 with strokes and 2 with tumors. They found that the mean frequency for key tapping averaged 3.5 Hz for paretic index fingers, similar to the 2.5 Hz value found with our dynamometer (Fig 3B⇑). Values for control subjects determined with the use of the present device also compare favorably with measurements made in other studies. Dominant hand index finger tapping by normal subjects in their seventh decade averages 3.6 Hz with a tally counter17 compared with 4.7 Hz in the present control group. The difference in tapping frequency may be due to the larger minimum force needed to produce a tap with the mechanical tally counter. The normalized value with the tally counter (0.95) is similar to the value (1.04) found in the present study. Dominant hand squeezing by normal subjects in their seventh decade averages 33.4 kg17 compared with 23.1 kg in the present control group. This difference may in part be attributable to use of a single grip size in the present study, since breadth of grip can influence force exerted.26 Despite these differences, the normalized value with a mechanical dynamometer (0.92) is similar to the value obtained in the present study (0.85).
When the motor performance of the unaffected hand was taken into consideration, the correlation of paretic hand dynamometer values with neurological scales improved. This improvement reflects a correction for the normal variability in motor performance seen among subjects.16,17,23 Prior reports of a small motor deficit in the unaffected hand after stroke8,18 suggest that use of the normalized value (paretic/unaffected hand) for dynamometer parameters in stroke subjects may slightly overestimate motor deficits. However, deficits in the unaffected hand after stroke may be least pronounced in tasks with little sensorimotor interaction,9 such as squeezing. Thus, although one prior study identified a deficit in squeezing force by the unaffected hand,8 three did not.7,9,18 Haaland,7 examining 43 patients with unilateral stroke, did not find a difference in index finger tapping frequency between the unaffected hand and control subjects. Indeed, in the present study no significant differences in tapping frequency or squeezing force were found between the unaffected hand and either control hand.
Dynamometer measurements of arm motor activities showed stronger correlations with the FM arm motor score than with the NIH stroke scale score. This pattern was expected, given the nature of each scale. The FM scale evaluates multiple motor functions, all involving the upper extremity. The NIH scale, however, includes points for abnormalities in a wide range of neurological modalities, including motor, sensory, language, vision, articulation, and attention. In addition, the NIH scale scores motor deficits in the lower extremities. The improved correlation and significance with use of the FM scale compared with the NIH stroke scale implies that abnormalities detected by the dynamometer are primarily arm motor deficits.
Dynamometer measurements of tapping and squeezing are linear measures that correlated closely with scores derived from valid, reliable scales. The findings suggest that in alert, cooperative stroke patients, the dynamometer is a reliable instrument for accurate assessment of arm motor status. The dynamometer allows for rapid assessment of simple and higher order motor parameters. This may be of value in studying and in measuring small changes in motor status during the period of stroke recovery. Future studies will clarify the value of serial dynamometer measurements during the stroke recovery process and evaluate the correlation of dynamometer measurements with scales in other neurological conditions.
This study was supported by a grant from the National Stroke Association (Dr Cramer). We are grateful to Pam Duncan, PhD, PT, for providing the training tape for the Fugl-Meyer scale. After completion of this study, a patent was filed on the described device by Drs Cramer and Finklestein. Thermocorps, Inc (Greenwood Village, Colo) has purchased the option to license the patent from Massachusetts General Hospital.
- Received May 9, 1997.
- Revision received August 28, 1997.
- Accepted August 28, 1997.
- Copyright © 1997 by American Heart Association
Gresham GE, Duncan PW, Stason WB, Adams HP, Adelman AM, Alexander DN, Bishop DS, Diller L, Donaldson NE, Granger CV, Holland AL, Kelly-Hayes M, McDowell FH, Myers L, Phipps MA, Roth EJ, Siebens HC, Tarvin GA, Trombly CA. Post-Stroke Rehabilitation. Rockville, Md: US Dept of Health and Human Services; 1995. Public Health Service, Agency for Health Care Policy and Research.
Muir KW, Weir CJ, Murray GD, Povey C, Lees KR. Comparison of neurological scales and scoring systems for acute stroke prognosis. Stroke. 1996;27:1817–1820.
Wood-Dauphinee SL, Williams JI, Shapiro SH. Examining outcome measures in a clinical study of stroke. Stroke. 1990;21:731–739.
Granger CV, Hamilton BB. Measurement of stroke rehabilitation outcome in the 1980s. Stroke. 1990;21(suppl II):II-46-II-47.
Dobkin BH. Neurologic Rehabilitation. Philadelphia, Pa: FA Davis; 1996.
Colebatch JG, Gandevia SC. The distribution of muscular weakness in upper motor neuron lesions affecting the arm. Brain. 1989;112:749–763.
Jones RD, Donaldson IM, Parkin PJ. Impairment and recovery of ipsilateral sensory-motor function following unilateral cerebral infarction. Brain. 1989;112:113–132.
Duncan PW, Propst M, Nelson SG. Reliability of the Fugl-Meyer assessment of sensorimotor recovery following cerebrovascular accident. Phys Ther. 1983;63:1606–1610.
Brott T, Adams HP, Olinger CP, Marler JR, Barsan WG, Biller J, Spilker J, Holleran R, Eberle R, Hertzberg V, Rorick M, Moomaw CJ, Walker M. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989;20:864–870.
Wityk RJ, Pessin MS, Kaplan RF, Caplan LR. Serial assessment of acute stroke using the NIH stroke scale. Stroke. 1994;25:362–365.
Fess EE. Grip strength. In: Casanova J, ed. Clinical Assessment Recommendations. 2nd ed. Chicago, Ill: American Society of Hand Therapists; 1992:41–45.
Trombly CA. Evaluation of biomechanical and physiological aspects of motor performance. In: Trombly CA, ed. Occupational Therapy for Physical Dysfunction. 4th ed. Baltimore, Md: Williams & Wilkins; 1995:143–156.
Spreen O, Strauss E. A Compendium of Neuropsychological Tests. New York, NY: Oxford University Press; 1991.
Desrosiers J, Bourbonnais D, Bravo G, Roy P-M, Guay M. Performance of the ’unaffected’ upper extremity of elderly stroke patients. Stroke. 1996;27:1564–1570.
Luce RD, Narens L. Measurement scales on the continuum. Science. 1987;236:1527–1532.
Hepp-Reymond MC, Wyuss UR, Anner R. Neuronal coding of static force in the primate motor cortex. J Physiol. 1978;74:287–291.
Evarts EV, Fromm C, Kroller J, Jennings VA. Motor cortex control of finely graded forces. J Neurophysiol. 1983;49:1199–1215.
Snow WG. Standardization of test administration and scoring criteria: some of the shortcomings of current practice with the Halstead-Rietan Test Battery. Clin Neuropsychol. 1987;1:250–262.
Chamorro A, Sacco RL, Mohr JP, Foukes MA, Kase CS, Tatemichi TK, Wolf PA, Price TR, Hier DB. Clinical-computed tomographic correlations of lacunar infarction in the Stroke Data Bank. Stroke. 1991;22:175–181.
Libman RB, Sacco RL, Shi T, Tatemichi TK, Mohr JP. Neurologic improvement in pure motor hemiparesis: implications for clinical trials. Neurology. 1992;42:1713–1716.
Harkonen R, Piirtomaa M, Alaranta H. Grip strength and hand position of the dynamometer in 204 Finnish adults. J Hand Surg (Br). 1993;18B:129–132.