Volume 87, Issue 9 , Pages 1207-1212, September 2006
Wavelet Analysis of Skin Blood Oscillations in Persons With Spinal Cord Injury and Able-Bodied Subjects
Article Outline
Abstract
Li Z, Leung JY, Tam EW, Mak AF. Wavelet analysis of skin blood oscillations in persons with spinal cord injury and able-bodied subjects.
Objective
To assess the blood oscillations in the skin over the ischial tuberosity (high-risk area for pressure ulcer) using spectral analysis of laser Doppler flowmetry signals based on wavelet transform.
Design
Wavelet analysis of skin blood oscillations in persons with spinal cord injury (SCI) and able-bodied subjects.
Setting
Seating and body support interface laboratory.
Participants
Ten men were recruited for this study, of whom 5 were able-bodied subjects (age, 31.2±3.3y) and 5 were persons with SCI (age, 37.2±7.3y).
Interventions
External pressure of 16.0kPa (120mmHg) was applied to the ischial tuberosity via 1 specifically designed pneumatic indentor. The loading duration was 30 minutes.
Main Outcome Measures
Skin blood flow was monitored for 10 minutes prior to loading and 20 minutes after the prescribed loading period. With spectral analysis based on wavelet transform, 5 frequency intervals were identified (.01–.02, .02–.06, .06–.15, .15–.40, .40–2.0Hz) corresponding to endothelial related metabolic, neurogenic, myogenic, respiratory, and cardiac activities, respectively.
Results
The relative amplitude of the metabolic component for persons with SCI was significantly lower (F=5.26, P=.032) during the resting conditions as compared with able-bodied subjects. During the postloading period, the response of oscillatory activities was evidently lower in the skin over the ischial tuberosity for persons with SCI when compared with able-bodied subjects. In addition, the relative amplitude of the neurogenic component (.02–.06Hz) during postloading was significantly lower for persons with SCI (F=5.44, P=.029).
Conclusions
These findings suggest that the contributions of endothelial related metabolic and neurogenic activities to the blood perfusion regulation become relatively less for persons with SCI during the resting and postloading periods, respectively.
Key Words: Laser-Doppler flowmetry , Pressure ulcer , Rehabilitation , Spinal cord injuries
TISSUE BREAKDOWN in the skin, leading to pressure ulcer formation, is a common complication developed in persons with spinal cord injury (SCI) when prolonged unrelieved pressure has been applied to the body and skin and underlying tissues. Pressure ulcer causation is a multifactorial process associated with a number of extrinsic factors (pressure, shear, duration of loading, temperature, moisture, hygiene)1, 2 and intrinsic factors (general physical conditions, morphology of bone and tissues, muscle tissue mechanical properties).3, 4 Although there are many factors contributing to the formation of pressure ulcers, a common pathway is associated with blood flow changes within the affected tissues.5 By evaluation of the skin microcirculation from the areas susceptible to pressure ulcer, Schubert et al6, 7 found a disturbed skin blood flow motion in persons with SCI during the resting state and at reactive hyperemia. However, the vascular response to pressure may be influenced by the effect of loss of neurogenic control and prolonged unrelieved pressure applied to body tissue after long-term paralysis. It is unclear whether the impaired autonomic nerve function in the patients with SCI contributes to a disturbed reactive hyperemic response.
Flow motion is the periodic oscillations of the blood flow in the vascular network which can be detected noninvasively using laser Doppler flowmetry (LDF).8, 9 Spectral analysis of the skin flow-motion signal deals with changes in the dynamics of blood flow and has been introduced as an approach for the evaluation of microvascular control mechanisms.10, 11, 12, 13 Analysis of these periodic oscillations may increase the knowledge of the dynamics of vascular control mechanisms.
Vascular flow-motion signals consist of notably different features in both time and frequency, of which high-frequency components have a shorter time span than the low-frequency components. To capture these features, methods in time-frequency domain analysis were used. The method of wavelet analysis has been applied to study the cardiovascular signals.14 Five characteristic frequencies of LDF signals have been identified in the human cutaneous circulation using wavelet analysis.15, 16 These oscillations reflect the influence of endothelial related metabolic activity, neurogenic activity on the vessel wall, intrinsic myogenic activity of the vascular smooth muscle, respiration, heart beat, with frequencies centered on .01, .04, 0.1, 0.3, and 1Hz, respectively.11, 12, 17, 18, 19
In persons with SCI, skin blood flow at the paralyzed tissue is generally disturbed. The specific origins of such disturbance have not been recognized but can be revealed using wavelet analysis of blood flow. Understanding the mechanisms of blood regulation between normal and paralyzed tissues may help to shed new insight on the possible causes of tissue breakdown, especially due to prolonged unrelieved pressure.
In the present study we assessed the peripheral vascular oscillations of tissue overlying the ischial tuberosity (high-risk area for pressure ulcer) in able-bodied persons and persons with SCI, using spectral analysis based on wavelet transform.
Methods
Participants
Ten men were recruited from a local medical rehabilitation center to participate in this study, of which 5 were able-bodied subjects (average age, 31.2±3.3y) and 5 were persons with SCI and complete paraplegia (average age, 37.2±7.3y). Three experimental trials were collected from each subject.
We asked all subjects if the following conditions were absent: (1) known case of vascular disease; (2) vasomotor drugs in use; (3) active pressure ulcer of any grade; (4) severe hypertonicity or remarkable contracture; (5) previous myocutaneous flap(s) or skin transplant surgery; and (6) heavy smoking or drinking. Heavy smoking was defined as the use of more than 2 cigarettes per day. Only 1 person, a subject with SCI (Pa_03) reported to be a light smoker (1–2 cigarettes per day). No alcohol drinks were permitted before the experiment. Table 1 shows the characteristics of the persons with SCI.
Table 1. Characteristics of Persons With SCI
| Subjects | Mode of Injury | Postinjury (y) | Level of Injury |
|---|---|---|---|
| Pa_01 | Traffic collision | 4 | T4 |
| Pa_02 | Fall injury | 7 | L2 |
| Pa_03 | Traffic collision | 12 | T7-8 |
| Pa_04 | Traffic collision | 10 | T5 |
| Pa_05 | Fall injury | 31 | T4 |
Before the experiment, basic parameters including the age, weight, height, and brief medical history about the subject were taken and a consent form was signed. The experimental protocol was reviewed and approved by the Hong Kong Polytechnic University Human Ethic Committee.
Experimental Procedure
All subjects were lying with both hips flexed to 90°. Figure 1 shows the experimental posture and setup. An LDF sensor was attached to skin over the ischial tuberosity by an adhesive ring. External pressure was applied to the skin by a pneumatic indentor (diameter, 25mm) aligned perpendicular to the loading site. A loading pressure of 16.0kPa (120mmHg) was chosen, based on Kernozek and Lewin,20 to simulate the interface pressure experienced by wheelchair users. The total duration of loading was 30 minutes.
Fig 1.
Schematic of experimental posture and setup. The testing posture was lying with both hips flexed to 90°. External pressure was applied to the skin by a pneumatic indentor (diameter, 25mm) aligned perpendicular to the skin over ischial tuberosity (IT). Both legs were supported by a height adjustable platform.
LDF Measurements
We performed blood flow measurements (in arbitrary units) using an LDF monitor (DTR4, software version 4.1)a with a contact probe (DP1T/7-V2),a with a power of 1.0mW at a wavelength of 780nm. The LDF signal was sampled at 40Hz and the measurements were expressed in arbitrary units in the tissue sample volume (flux) for comparative study. The calibration of the laser Doppler flowmeter was performed regularly using a standard reference (flux standard) provided by the manufacturer. The standard uses the brownian movement of polystyrene microspheres in water to produce the reference signals. Skin blood flow was monitored for 10 minutes prior to loading and two consecutive 10-minute intervals (t1, t2) after the release of loading.
Spectral Analysis
We commonly performed spectral analyses using Fourier or short-time Fourier methods. However, Fourier transform analysis does not provide sufficient time resolution for analysis of nonstationary physiologic signals, such as blood flow signal.18 In the blood perfusion signal, both low- and high-frequency components are present and a Fourier transformation fails either in the following of the time evolution of the high-frequency events or in the estimation of the frequency content of the low-frequency band. The method of wavelet analysis offers a solution to this problem.
The idea of the continuous wavelet transform is to project a signal s on a family of zero-mean functions, the wavelets, deduced from an elementary function, called the mother wavelet Ψ(u), by translations and dilations. Continuous wavelet transform of a signal s is defined as
(1)
(2)The continuous wavelet transform is a mapping of the function g(u) onto the time-frequency plane. By adjusting the window used in wavelet transform, slower and faster events can be categorized accordingly.21 In this study, Morlet wavelet was chosen for the wavelet transform analysis. Morlet wavelet is a Gaussian function that allows the best time-frequency localization within the limits given by the uncertainty principle.14, 18, 21, 22
The wavelet transform was calculated in the frequency interval from 0.01 to 2Hz. Typical wavelet transform of LDF signal in (1) time-frequency plane and (2) averaged over time, are presented in figure 2. The wavelet transform is calculated with a logarithmic resolution, and the frequency axes are therefore presented logarithmically. This is particularly appropriate for the estimation of the low-frequency component. In this interval, periodic oscillations with 5 characteristic frequencies (centered on .015, .03, 0.1, 0.3, and 1Hz, respectively) were observed. The outer limits of each spectral interval were determined: (1) from .01 to .02Hz, (2) from .02 to .06 Hz, (3) from .06 to .15Hz, (4) from .15 to .4Hz, and (5) from 0.4 to 2Hz. All mathematical algorithms were developed using Matlab, version 6.5.b
Fig 2.
The wavelet transform in (A) time-frequency plane and (B) averaged spectrum of LDF signal shown on a log scale. The vertical lines indicate the outer limits of frequency interval: (1) .01 to .02Hz, (2) .02 to .06Hz, (3) .06 to .15Hz, (4) .15 to 0.4Hz, and (5) 0.4 to 2Hz, which correspond to endothelial related metabolic, neurogenic, myogenic, respiratory and cardiac activities, respectively.
Quantitative Measures
An oscillatory component in a signal can be characterized by its instantaneous frequency and corresponding amplitude. Quantitative measures were introduced to make comparisons between sets of signals.14 The average amplitude within each interval was used to characterize the spectral components. The average amplitude of a spectral component in a given frequency interval can be determined as
(3)
(4)Statistical Analysis
We expressed all values as means and standard deviations. For intersubject comparison, the absolute amplitudes of blood oscillations were normalized to the baseline skin blood flowmotion at the time of preloading and postloading. The data of the subject were tested for the normality (Kolmogorov-Smirnov test) and homogeneity of variance (Levene test) to ensure they meet the assumption required by the parameter analysis. We used the t test to test the significant difference in age and body mass index (BMI) between able-bodied control subjects and persons with SCI. We used one-way repeated-measures analysis of variance (ANOVA) to study the differences in the blood oscillations for each frequency band between preloading and postloading periods in persons with SCI or able-bodied subjects. We used two-way repeated-measures ANOVA to study the significant difference in the relative amplitude of blood oscillations for each frequency band between preloading and postloading periods in persons with SCI and able-bodied subjects.c Post hoc analyses within the groups were done with Bonferroni multiple comparison tests. A difference with P less than .05 was considered statistically significant.
Results
Participant Characteristics
The BMI was 25.2±3.0kg/m2 and 22.5±5.8kg/m2, respectively, in the able-bodied group and the SCI group. No significant differences was found in age (P=.13) and BMI (P=.38) in the 2 groups.
Response of Blood Oscillatory Activities to the Locally Applied Pressure on Persons With SCI and Able-Bodied Subjects
Figures 3A and B show the normalized absolute and relative amplitudes for the 5 frequency intervals for the resting skin over the ischial tuberosity in the able-bodied subjects during the preloading and postloading periods. One-way repeated-measures ANOVA showed that the absolute oscillatory activities were significantly higher by 114% in interval 1, 132% in interval 2, 45% in interval 3, 130% in interval 4, and 66% in interval 5 following the release of loading during the first postloading period (t1) as compared with that during the preloading period and these oscillatory activities returned to the preloading level during the second postloading period (t2) (P<.05 for the metabolic component [interval 1, .01–.02]; P<.01 for other components) (see fig 3A). No significant change in the relative amplitude was found except the myogenic component (interval 3, .06–.15), in which the relative amplitude was significantly lower by 30% during the first postloading period (t1) as compared with that during the preloading period (P<.05) and increased significantly during the second postloading period (t2) (P<.01) (see fig 3B).
Fig 3.
Normalized absolute and relative amplitudes for the 5 frequency intervals for the resting skin over the ischial tuberosity in (A, B) the able-bodied subjects and (C, D) persons with SCI during the preloading and the 2 postloading periods (t1, t2). Significant differences are marked between postloading (t1) and preloading with *P<.05 and †P<.01; postloading (t2) and postloading (t1) with ‡P<.05 and §P<.01. Each value was calculated as a percentage of each PRE value in the same group.
Figures 3C and 3D show the normalized absolute and relative amplitudes for the 5 frequency intervals for the resting skin over the ischial tuberosity in persons with SCI during the preloading and postloading periods. Only the absolute activities in interval 1 (.01–.02Hz) was found to be significantly higher, by 88%, for the subjects with SCI during t1 (P<.05) and this activity returned to the preloading levels during t2 (P<.01) (see fig 3C); whereas the relative activities in intervals 2 (.02–.06Hz), 3 (.06–.15Hz), and 4 (.15–.4Hz) were significantly lower by 25%, 30%, and 40%, respectively, during t1 as compared with the preloading period and returned to the preloading level during t2 (interval 2, P<.05; intervals 3, 4, P<.01) (see fig 3D).
Comparison of Relative Oscillatory Activities Between the Able-Bodied Subjects and Persons With SCI
The relative amplitude of the metabolic component (interval 1, .01–.02) for the subjects with SCI was found to be significantly lower by 20% during the preloading period when compared with able-bodied subjects (F=5.26, P=.03) (fig 4A). Also the relative amplitude in interval 2 (.02–.06Hz) was significantly lower by 27% for subjects with SCI during t1 when compared with able-bodied subjects (F=5.44, P=.03) (fig 4B). Typical examples of time-averaged wavelet transform calculated from signals measured in the resting skin over the ischial tuberosity in (A) an able-bodied subject and (B) a person with SCI are presented in figure 5.
Fig 4.
Relative amplitude for the 5 frequency intervals for the resting skin over the ischial tuberosity in the able-bodied subjects and persons with SCI during the (A) preloading and (B) postloading periods (t1). *Significant differences are marked between persons with SCI and able-bodied subjects (P<.05).
Fig 5.
Typical examples of time-averaged wavelet transform calculated from signals measured in the resting skin over the ischial tuberosity in (A) an able-bodied subject and (B) a person with SCI. The vertical lines indicate the outer limits of each frequency interval from 1 to 5, which correspond to endothelial related metabolic, neurogenic, myogenic, heart, and respiratory activities, respectively.
Discussion
The first difference in the oscillatory activities between persons with SCI and able-bodied subjects was that the endothelial related relative metabolic activity (interval 1, .01–.02Hz) was significantly lower during the preloading period for persons with SCI compared with that for the able-bodied subjects. The most important role of such metabolic regulation is to adjust the blood flow to satisfy the need of cells for oxygen.23 Such metabolic activity apparently originates from the endothelial layer.21 In our previous study, it has been shown that prolonged surface loading induced significant decrease in the relative activity of the metabolic component (.01–.04Hz) at the trochanter area in the anesthetized rat.24 The lower metabolic activity in the skin over the ischial tuberosity area might be attributed to prolonged unrelieved pressure for persons with SCI. The fact that the relative amplitude of the oscillations of this frequency interval was lower indicates that the endothelial related metabolic activity might contribute relatively less to the regulation of blood perfusion in persons with SCI at rest. This may result in a reduced oxygen delivery due to an increase in vascular resistance.
Another significant difference in the oscillatory activities between persons with SCI and able-bodied subjects was evident in the response of the oscillatory activities following the release of loading in the skin over the ischial tuberosity area. After pressure loading for 30 minutes, all absolute oscillatory activities reached a significantly higher level and then returned to the preloading level for the able-bodied subjects. However, for persons with SCI, only the absolute metabolic activity (interval 1, .01–.02) showed the same postocclusive response compared with the able-bodied subjects. In fact, decreased relative amplitudes in the frequency intervals 2 (.02–.06Hz), 3 (.06–.15Hz), and 4 (.15–0.4Hz) (corresponding to neurogenic, myogenic, and respiratory activities, respectively) were found in persons with SCI during t1. The reduced reactive blood oscillation activity in the skin over the ischial tuberosity area reported here is supported by the study of Schubert6 and Hagisawa25 and colleagues, who reported a lower increase of the skin blood flow or a lower immediate response of blood content to load release over a risk area for pressure ulcer in persons with SCI during postocclusive hyperemia response.
Reactive hyperemia is a consequence of vasodilation following occlusion of arteriole.26, 27, 28 Vasodilation mechanism involves the stimuli resulting from metabolic changes subjected to hypoxia and accumulation of metabolites,29, 30, 31 the release of mediators from the endothelium or from afferent nerves.28, 32, 33 Lower relative amplitude of the neurogenic component (.02–.06Hz) found in persons with SCI indicated that the loss of neurogenic control might be a main factor contributing to the disturbance in the postpressure regulation of the oscillatory flow. The neurogenic regulation is based on the global integration of the neural signals. It permits the body to react to its environment and to make adjustments based on internal and external factors. It is clear from the present study that the ability of the tissue to provide the large reactive flow motion response in the skin over the ischial tuberosity area was significantly impaired in persons with SCI. The hyperemic response following the release of occlusive pressure represents a protective adaptation to the damaging effects of pressure on the skin.34 Failure of this protective mechanism may lead to insufficient skin recovery following the release of occlusive pressure.
Decreased relative amplitudes in the frequency interval 3 (.06–.15Hz) was also found during t1 for the able-bodied subjects. Local myogenic control is considered independent of any neural or humoral influences and has been associated with the application of mechanical force to the vascular smooth muscle cell.35, 36 The applied pressure might generate substantial stress to deform the underlying arterioles and affect the wall stress of the smooth muscle layer of the arteriolar, which significantly affects contractility.37, 38
A limitation of the current study was the small sample size used. Although our results demonstrate that the oscillatory activities were different between persons with SCI and able-bodied subjects, more work is still required to substantiate the generalization of these findings across persons with SCI. The disturbed flow motion in persons with SCI may have an important effect on oxygen delivery to tissue that could be an important precursor to tissue breakdown. This will need to be clarified in future studies.
Conclusions
In the absence of neurogenic control for persons with SCI, the lower relative metabolic activity in the skin over the ischial tuberosity area at rest might be the result of prolonged unrelieved pressure applied to the body and skin and underlying tissues compared with the able-bodied subjects. However, the reduced response of oscillatory activities following the release of occlusive pressure might be mainly attributed to the loss of neurogenic control. These findings suggest that the contributions of endothelial related metabolic and neurogenic activities to the blood perfusion regulation become relatively less for persons with SCI during the resting and postloading periods, respectively. This may result in reduced oxygen delivery and insufficient skin recovery.
Suppliers
References
- . Effects of compressive load regimens on tissue viability . In: Bader DL editors. Pressure ulcers (clinical practice and scientific approach) . Basingstoke: Macmillan; 1990;p. 191–201
- . Establishing predictive indicators for the status of loaded soft tissues . J Appl Physiol . 2001;90:2231–2237
- . In vitro models to study compressive strain-induced muscle cell damage . Biorheology . 2003;40:383–388
- . Mechanical compression-induced pressure sores in rat hindlimb (muscle stiffness, histology, and computational models) . J Appl Physiol . 2004;96:2034–2049
- . Effects of different cyclic pressurization and relief patterns on heel skin blood perfusion . Adv Skin Wound Care . 2002;15:158–164
- . Postocclusive reactive hyperemia and thermal response in the skin microcirculation of subjects with spinal-cord injury . Scand J Rehabil Med . 1991;23:33–40
- . Analysis of arterial flowmotion in spinal-cord injured and elderly subjects in an area at risk for the development of pressure ulcers . Paraplegia . 1995;33:387–397
- . Variations in laser Doppler flux and flow motion patterns in the dorsal skin of the human foot . Microvasc Res . 1990;39:212–222
- . Evaluation of flowmotion in man by laser-Doppler technique . Blood Vessels . 1991;28(Suppl 1):21–26
- . Is high-frequency fluxmotion due to respiration or to vasomotion activity? . In: Messmer K editors. Progress in applied microcirculation . Basel: Karger; 1993;p. 52–58
- . Dynamic coherence analysis of vasomotion and flow motion in skeletal muscle microcirculation . Microvasc Res . 1996;52:235–244
- . Respiratory-dependent laser-Doppler flux motion in different skin areas and its meaning to autonomic nervous control of the vessels of the skin . Microvasc Res . 1996;52:69–78
- . A comparison of changes in rhythms of sacral skin blood flow in response to heating and indentation . Arch Phys Med Rehabil . 2005;86:1245–1251
- . Wavelet based analysis of human blood flow dynamics . Bull Math Biol . 1998;60:919–935
- . Spectral analysis of the laser Doppler perfusion signal in human skin before and after exercise . Microvasc Res . 1998;56:173–182
- . Using wavelet analysis to characterize the thermoregulatory mechanisms of sacral skin blood flow . J Rehabil Res Dev . 2004;41:797–806
- . Vasomotion in human skin before and after local heating recorded with laser Doppler flowmetry . Int J Microcirc Clin Exp . 1989;8:205–215
- . Wavelet analysis of oscillations in the peripheral blood circulation measured by laser Doppler technique . IEEE Trans Biomed Eng . 1999;46:1230–1239
- . Involvement of sympathetic nerve activity in skin blood flow oscillations in humans . Am J Physiol Heart Circ Physiol . 2003;284:H1638–H1646
- . Seat interface pressures of individuals with paraplegia (influence of dynamic wheelchair locomotion compared with static seated measurements) . Arch Phys Med Rehabil . 1998;79:313–316
- . Oscillations in the human cutaneous blood perfusion signal modified by endothelium-dependent and endothelium-independent vasodilators . Microvasc Res . 1999;57:298–309
- . Sampling theory and wave propagation . In: Chen CH editors. NATO ASI Series: Issues in acoustic signal/image processing and recognition. Vol 1 . Berlin: Springer Verlag; 1983;p. 233–261
- . Time-frequency analysis of laser Doppler flowmetry signals recorded in response to a progressive pressure applied locally on anaesthetized healthy rats . Physics Med Biol . 2004;49:843–857
- . Effects of prolonged surface pressure to the skin blood flowmotions in anaesthetized rats—assessment by spectral analysis of laser Doppler flowmetry signals . Phys Med Biol . 2006;51:2681–2694
- . Assessment of skin blood content and oxygenation in spinal-cord injured subjects during reactive hyperemia . J Rehabil Res Dev . 1994;31:1–14
- . The recovery characteristics of soft tissues following repeated loading . J Rehabil Res Dev . 1990;27:141–150
- . A comparison of ischemic and pressure induced hyperemia . Arch Phys Med Rehabil . 1998;79:1451–1455
- . Mechanisms of the cutaneous vasodilator response to local external pressure application in rats (involvement of CGRP, neurokinins, prostaglandins and NO) . Br J Pharmacol . 2000;131:1161–1171
- . A peripheral arterial conducting mechanism underlying dilatation of the femoral artery and concerned in functional vasodilatation in skeletal muscle . J Physiol . 1959;149:93–111
- . Coordination of flow control among microvessels by intercellular conduction . Science . 1986;234:868–870
- . Communication between feed arteries and microvessels in hamster striated muscle (segmental vascular responses are functionally coordinated) . Circ Res . 1986;59:283–290
- . Antagonism of acetylcholine and adenosine rat cremaster arteriolar vasodilation by combination of NO antagonists . Int J Microcirc Clin Exp . 1993;12:275–286
- . Non-nociceptive capsaicin-sensitive nerve terminal stimulation allows for an original vasodilatory reflex in the human skin . Brain Res . 1998;811:166–168
- Post pressure hyperemia in the rat . Comp Biochem Physiol A Mol Integr Physiol . 2004;137:533–546
- . Cellular mechanisms involved in the vascular myogenic response . Am J Physiol . 1992;263(3 Pt 2):H647–H659
- . Intraluminal pressure modulates the magnitude and the frequency of induced vasomotion in rat arteries . J Vasc Res . 1995;32:237–246
- . Effect of stretch on passive tension and contractility of isolated vascular smooth muscle . Am J Physiol . 1962;202:835–840
- . Wall stress (a determinant of regional differences in response of frog microvessels to norepinephrine) . Am J Physiol . 1972;222:82–91
Supported by the Central Research Grants of the Hong Kong Polytechnic University (grant nos. G-YD72, G-YX35).No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated.
PII: S0003-9993(06)00522-3
doi:10.1016/j.apmr.2006.05.025
© 2006 American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation. Published by Elsevier Inc. All rights reserved.
Volume 87, Issue 9 , Pages 1207-1212, September 2006
