201 posters, 
1 session, 
11 topics, 
149 authors
0 institutions
ePostersOnline®, by the ePostersLive® team

121

Real-Time Detection of Spinal Cord Position for Adaptive Stimulation using Near Infrared Reflectometry


Rating
Your rating: None Average: 2 (3 votes)
Part of Session

Session

Part of Topic

Spinal Cord Stimulation


Real-Time Detection of Spinal Cord Position for Adaptive Stimulation using Near Infrared Reflectometry

Erich W. Wolf,II, MD, PhD, Center for Neurosurgical and Spinal Disorders, Lake Charles, LA

ABSTRACT:

Introduction:  Motion of the spinal cord relative to the epidural electrode array can cause suboptimal stimulation: either noxious, inefficient, or insufficient. Adaptive stimulation attempts to mitigate these effects by modulating stimulation parameters in a position-dependent fashion. Near-infrared (NIR) reflectometry is demonstrated to provide real-time direct measurement of spinal cord position at the site of stimulation which can facilitate closed-loop adaptive stimulation during static and dynamic motion states.  Methods: A miniature sensor array consisting of a NIR light emitting diode flanked by phototransistors potted in epoxy was placed in the dorsal epidural space of a human cadaver at the T8 level via laminotomy. Turgor of the subarachnoid space was maintained by continuous intrathecal infusion of normal saline via catheter at the craniocervical junction. NIR reflectance was measured as the cadaver was rotated about its longitudinal axis on a gantry. 16-bit analog-to-digital converters acquired stereoscopic reflectance data and angular position of the gantry was measured using a coaxial multi-turn potentiometer at a 10-Hz sampling rate. Gantry angular velocity was computed from the first derivative of the gantry position data. Data analysis was performed using IgorPro software.  Results: Results from near-infrared reflectometry detection of spinal cord position are consistent with anticipated cord position based upon prior imaging studies in static, ordinal patient positions. Gravitational force was the primary determinant of cord position in the static or quasi-static sense. Under dynamic motion conditions there was statistically-significant (Pearson's R=0.58) cross-correlation between reflectometry data and the tangential velocity-squared, suggesting that centripetal force was the primary determinant of cord position as the gantry was rotated. Decomposition of the reflectometry data by subtracting out the centripetal force component revealed the expected gravitational force component.  Conclusions: Spinal cord position during dynamic motion has been shown to differ from static predictions due to additional influences such as centripetal force. These findings underscore limitations in extrapolating spinal cord position from surrogates such as body position or body acceleration at sites remote from the stimulating electrodes. NIR reflectometry offers a real-time direct measure of spinal cord position in both static and dynamic motion states for closed-loop adaptive stimulation applications.
Introduction:
Spinal cord stimulation (SCS) is clinically employed for the treatment of chronic pain in carefully selected patients.  An implanted generator delivers a train of electrical pulses to an epidural electrode array positioned in the dorsal spinal canal, usually within the thoracic region.  An induced electric field within the dorsal columns of the spinal cord produces a tingling paresthesia which interferes with pain perception.  For optimal pain relief the paresthesias must somatotopically cover the painful regions without themselves being noxious.  The somatotopic distribution is dictated by the position of target neurons within the spinal cord relative to anodic and cathodic contacts within the epidural electrode array and pulse-width.  Intensity of the paresthesias is dictated by a combination of stimulus amplitude, pulse-width, and frequency.
Clinical observations show these paresthesias to be subject to changing magnitude and topographical distribution during patient positional changes given static stimulation parameters.  This is thought to result from migration of the spinal cord in the axial plane relative to the electrode array.  Holsheimer, et al. in an in vivo MRI study have demonstrated spinal cord translation of several millimeters in a gravity-dependent fashion within the spinal canal.  Subsequent finite-element analysis by Molnar, et al. showed significant changes in dorsal column activation with positional changes but without appreciable changes in impedance.  
To mitigate unwanted paresthetic perturbations, the induced electric field must be modulated to accommodate positional changes.  Schade, et al. reported favorable clinical results using an accelerometer-based system which was able to resolve six discrete patient activities or positions: lying supine, lying right lateral, lying left lateral, lying prone, standing upright, and walking.  The accelerometer data was used to select between predefined programs for which the stimulation had been optimized for each position or activity.  While this technology proved clinically advantageous, there remain limitations imposed by using body position as a surrogate for spinal cord position, discretization of body position, absence of stimulation parameter interpolation between positions, and sensor dwell-time.
To achieve fully adaptive stimulation it is necessary to ascertain actual spinal cord position relative to the electrode array in real-time to provide dynamic control of the electric field. This may be realized using near-infrared (NIR) reflectometry.  
Body tissues are relatively translucent at the near-infrared light spectrum.  An NIR emitter situated along the sagittal plane of the spinal canal in the epidural space illuminates the spinal cord.  NIR light easily passes through the dura and cerebrospinal fluid to reflect off the surface of the spinal cord.  Reflected NIR light traverses back through the dura to be detected by stereoscopic photodetectors situated on either side of the emitter.  The magnitude of reflected NIR light is inversely proportional to distance between the spinal cord and detectors.  The sagittal distance between spinal cord and sensor array (comprised of the emitter and paired detectors) may be ascertained by the summed photodetector output.  The coronal displacement of the spinal cord may be determined by the difference signal.  
Methods:
Sensor Fabrication:
The sensor array comprised two OP501A phototransistors straddling a Vishay VSMY1850CT light-emitting diode (LED) as shown in Figure 1.  These 0805 format surface-mount devices (SMD) were soldered onto hookup wire arranged in co-planar fashion such that the leads projected perpendicularly to the axis of the linear array. The array was then potted in EPO-TEK 301-2 (Epoxy Technology, Inc., Billerica, MA) epoxy.
Cadaver Preparation:
An adult human cadaver with intact craniospinal axis was prepared by insertion of a 14-gauge catheter dorsally at the occipitocervical junction.  Lactated Ringer’s solution was infused using a pressure bag to simulate cerebrospinal fluid at the appropriate thecal sac turgor.  An open T8 midline laminotomy was performed with resection of the ligamentum flavum, epidural fat, and epidural veins to reveal the dura.  The sensor array was then placed in the laminotomy defect against the dura as shown in Figure 2.  The cadaver preparation was placed longitudinally in a custom-made rotating gimbal shown in Figure 3.  A polar coordinate system was defined relative to the gimbal longitudinal axis such that the cadaver was in the prone position at zero degrees.  The gimbal was angled approximately 15 degrees Trendelenberg to prevent subdural air bubbles from accumulating within the thoracic spine.  Angular position of the gimbal was measured by displacement of a 10-turn potentiometer (Milipot RV-10K, Oak Engineering, Oakland, CA.
Data Acquisition and Processing:
Right and left photodetector outputs, rotatory position sensor, and  LED forward bias voltage were measured using a National Instruments (Austin, TX) USB-6210 16-bit  single-ended analog-to-digital converter at 10- to 60-Hz sampling rates using NI LabView SignalExpress.  Data from SignalExpress log files were imported to Microsoft (Seattle, WA) Excel software then saved to .xls format for cross-platform importation to IGOR Pro V6.31 software (WaveMetrics, Inc., Oswego, OR) running on Apple, Inc. (Cupertino, CA) OS X 10.8.3.  Data analysis and graphing was performed within IGOR Pro.
RESULTS:
Photodetector Quasi-Static Response:
Figure 4 demonstrates the system response to 90-degree step-wise decrements of gimbal rotation.  Both left and right photodetector outputs respond in synchronous fashion to positional change with local maxima at 180-degrees modulo 360 degrees.  The latter is consistent with increased NIR reflection in the supine position.  
A subset of data from Figure 4 (720- to 360-degrees modulo 360) is graphed in Figure 5 which demonstrates the symmetry about the 180-degree point.  These data, reflected about 180-degrees, may be closely described by Gaussian, third-order polynomial, sigmoidal, and sinusoidal curve fits.   
 Photodetector Dynamic Response:
The sensor dynamic response was evaluated by rapidly turning the gimbal forward and reverse by one full rotation repeatedly.  The sum of the photodetector outputs during this motion is shown in Figure 6.  The data of Figure 6 was decomposed into six wavelets (labeled) spanning the rising slope of the gimbal angular rotation.  Each wavelet amplitude was normalized to remove DC-offset and baseline drift (introduced by leakage currents) over the wavelet period.  Resultant wavelets shown in Figure 7 have a characteristic, repeatable pattern but differ from the expected result.  Specifically, the anticipated local maxima are expected at 180-degrees based upon the findings of Figure 4.  Furthermore, the phase shift between tracings of Figure 7 were found to correlate with gantry angular velocity.  These data suggested  that spinal cord position was influenced by mechanical variables.  
Considering a free-body diagram of a spinal cord which is rotating about an axis, the positional dependence upon angular velocity suggested that centripetal force was the primary determinant of position in the dynamic context.  This is in contradistinction with the quasi-static experiments in which the gravitational force predominated.  
The summed photodetector output was then plotted with tangential velocity-squared as shown in the time-scale magnification of Figure 8.  The cross-correlation of these data demonstrated statistical significance at p < 0.05 using Fisher's z-transformation with Pearson’s correlation (r = 0.58).
These data suggest that spinal cord position may be predicted from body position in the static- or quasi-static sense but not necessarily in the dynamic state.  In the static or quasi-static condition the spinal cord position is dictated by gravitational force checked by suspensory forces offered by the paired nerve roots, denticulate ligaments, and arachnoid.  However, in the dynamic context, centripetal force, angular acceleration, linear acceleration, and cerebrospinal fluid viscosity may become influential.  Decomposition of the net force acting on the spinal cord by removing the centripetal force would be expected to unmask the effects of the gravitational force.  This is demonstrated in Figure 9.  Unity-normalized summed photodetector output from wavelet 6 (bottom trace) was subtracted from similarly normalized tangential velocity-squared (middle trace) to arrive at the the top trace.  The effect of the gravitational force is elucidated at the expected 180-degrees.  
 
Discussion:
Near-infrared reflectometry has been demonstrated to provide a real-time indication of spinal cord position within the spinal canal.  A dynamically-adaptive SCS system may be fashioned by incorporating an NIR-reflectometer into the epidural electrode array and utilizing software algorithms to interpolate stimulation parameters.
Of primary importance is knowing spinal cord position in the sagittal plane.  This information facilitates modulation of the electric field amplitude, thus reducing noxious stimuli and inadequate stimuli.  These data may be gleaned with a minimum hardware requirement of one LED and one photodetector although using the summed outputs of stereoscopic detectors have a theoretical advantage in minimizing interdependence of coronal translation of the spinal cord.  Further, the difference signal between stereoscopic detectors may offer information regarding coronal displacement.
The work of Holsheimer, et al., serves as a gold-standard for relative cord position within the spinal canal for certain static, ordinal body positions.  In the experiments detailed here, there was excellent correlation between the summed photodetector outputs and anticipated spinal cord position in the quasi-static sense as shown in Figure 4.  No data exists within the literature regarding spinal cord position during dynamic patient motion.  The absence of a gold-standard for spinal cord position during motion complicated the data analysis as the dynamic system response differed from that expected in the static state.  
Analysis of the system dynamic response revealed a statistically-significant cross-correlation between the photodetector summed output and the square of the tangential velocity.  In a physical context, this suggested that the centripetal force, caused by off-axis rotation of the spinal canal about the gimbal axis, influenced spinal cord position within the canal more than the gravitational-dependent findings of Holsheimer. The ability of NIR-reflectometry to detect spinal cord position in both static and dynamic states may supplant our prior gold-standard.  This capability facilitates the implementation of a position-adaptive and activity-adaptive spinal cord stimulation system.  Cervical spinal cord stimulation for upper extremity pain, now limited in practice by cervical cord motion, may find much wider adoption.  Further, NIR-reflectometry may find utility in applications such as brain injury research and correction of brain shift during stereotaxy.
Several curve-fits modeled the data with sufficient fidelity to serve as a basis for interpolation algorithms.  In a practical implementation, stimulation parameters would be manually optimized in each of four static, ordinal positions:  laying prone, laying supine, left lateral and right lateral positions.  These positions represent the extremes of spinal cord position due to gravitational force, but not necessarily with a superimposed acceleration.  Modulation of the stimulation parameters may be dictated by interpolation within, or by extrapolation outside, the confines of the cord position associated with these static ordinal positions through the use of a curve-fit.
Conclusions:
Results from near-infrared reflectometry detection of spinal cord position are consistent with anticipated cord position based upon prior imaging studies in static, ordinal patient positions.  The ability to transduce position ad lib and at high repetition rates offers functionality in both quasi-static and dynamic motion states.  Spinal cord position during dynamic motion has been shown to differ from static predictions due to additional influences such as centripetal force.  These findings underscore limitations in extrapolating spinal cord position from surrogates such as body position or body acceleration at sites remote from the stimulating electrodes.  Real-time NIR data may be used to modulate stimulation parameters to achieve both position-adaptive and activity-adaptive stimulation.  Reproducible position data which is concordant with several curve-fits suggest that stimulation parameters may be interpolated using parameters optimized for four static, ordinal positions of the cord within the spinal canal.  This may be achieved through programming with the patient in prone, supine, right lateral, and left lateral positions.
Incorporating near-infrared reflectometry into implanted spinal cord stimulation systems is anticipated to mitigate the intermittent noxious paresthesiae associated with gravity-dependent position and dynamic motion.
 
REFERENCES: Holsheimer J, den Boer JA, Struijk JJ, Rozeboom AR. MR assessment of the normal position of the spinal cord in the spinal canal.  Am J Neuroradiol 15(5): 951-9, 1994. 
Molnar G, Panken C, Kelley K. Effects of spinal cord movement and position changes on neural activation patterns during spinal cord stimulation. Abstract. American Academy of Pain Medicine. San Antonio, TX: Feb. 3-6, 2010.
Schade, CM, Schultz, D Tamay, N, Iyer, S, Panken, E.  Automatic Adaptation of Neurostimulation Therapy in Response to Changes in Patient Position: Results of the Posture Responsive Spinal Cord Stimulation (PRS) Research Study.  Pain Physician 14:407-17, 2011.