Identification of the Epidural Space with Optical Spectroscopy: AnIn Vivo Swine Study

• ❖ Identification of the epidural space during needle insertion relies on feel, which can be misleading.

• ❖ In one anesthetized pig, optical spectra at the epidural needle tip displayed a large increase in lipid volume fraction upon entering the epidural space.

• ❖ Optical imaging may provide direct, nontactile feedback of epidural space entry with a needle.

EPIDURAL injection is a common intervention for providing surgical anesthesia as well as for treating acute radicular pain associated with lumbar intervertebral disc herniation.1Accurately identifying the epidural space during these procedures can be challenging. The loss-of-resistance (LOR) technique is most frequently used for identification, but in the absence of confirmatory methods, it is often inaccurate. It has been estimated that the LOR technique is inaccurate in up to 53% of cervical epidural steroid injections when used without image guidance.2Failure of the LOR can occur when dense ligamentum flavum is absent, so that significant resistance to injection is not encountered.3 

Complications associated with epidural injections have been well documented.1The most frequent complication is placement of the injected solution in a location outside of the epidural space, which is likely to result in inadequate anesthesia for surgical procedures (epidural anesthesia) and to reduce the efficacy of treatment for radicular pain (epidural steroid injections). Dural puncture can result in postdural puncture headache and placement of solutions within the cerebrospinal fluid, resulting in spinal rather than epidural anesthesia; the safety of the excipient containing the steroid in this context is uncertain.4Intravascular injections of local anesthetic or particulate steroids, which can occur even when aspiration is negative for blood,5can result in cardiorespiratory arrest, central nervous system toxicity, and ischemic neurologic events in the spinal cord and brain.1Spinal cord damage6and paralysis after injection within the spinal cord7have also been reported.

Fluoroscopic guidance has resulted in significant improvements in the accuracy of needle tip localization and has reduced complication rates; as such, it is strongly recommended for patients receiving epidural steroid injections.8A recent study found that intrathecal needle placement still occurred in 0.5% of interlaminar lumbar epidural steroid injection procedures in which both fluoroscopy and the LOR technique were used.9Fluoroscopic guidance remains impractical for routine use in surgical and obstetric anesthesia.

Recently, optical identification of the epidural space with a porcine model was demonstrated.10In that study, a custom stylet with integrated optical fibers, positioned in a Tuohy needle, was used to deliver and receive light at two different wavelengths in the visible range. Although the results are very promising, several important questions remained unaddressed. First, how do optical signals acquired from the needle correspond to physiologic parameters of tissue? Second, can adipose tissue present in the epidural space be identified using measurements in the near-infrared range, where optical contrast for lipids is particularly high and not confounded by the presence of hemoglobin? Third, can optical signals be used to identify intravascular and intrathecal needle placement? The present study, designed and conducted without knowledge of the study of Ting et al .,10addresses these questions. A custom optical spinal needle designed for practical use during regional anesthesia was used. It has optical fibers embedded in the cannula that allow for optical spectroscopic measurements to be acquired from tissues close to the bevel surface of the needle tip. We performed insertions to the epidural space of a swine in vivo , with both midline and paramedian approaches. During each insertion, we obtained optical spectra spanning visible and near-infrared wavelengths at multiple insertion depths. Differences in optical spectra associated with the corresponding tissue transitions were quantified with a spectral processing algorithm that derived estimates of the blood and lipid volume fractions in tissues close to the bevel surface of the needle tip.

Experiments were performed at Karolinska University Hospital with Internal Review Board approval (Northern Regional Ethics Committee for Animal Research, Stockholm, Sweden). Given general anesthesia, a Swedish Landrace swine (42 kg; male) was positioned on a fluoroscopy table in the left lateral decubitus position. General anesthesia was induced with sodium thiopental and maintained with sodium thiopental and fentanyl infusions throughout the experiment. Mechanical ventilation and continuous monitoring of heart rate and pulse oximetry were performed.

The design of the optical spinal needle was derived from an 18-gauge spinal injection needle with a Quincke model tip. Three optical fibers (core diameter: 200 μm) were embedded in the needle shaft, so that they terminated at the bevel surface. One fiber was located near the distal end of the bevel surface and delivered broadband light, which was scattered and absorbed by the tissue in front of the needle tip. The two other optical fibers were located at the proximal end of the bevel surface, at a distance of 2.4 mm from the distal fiber; they received light scattered within tissue and directed it to two spectrometers where it was spectrally resolved across the wavelength range of 500–1,600 nm. From each spectrum that was acquired from tissue, a background spectrum that was acquired with the broadband light shuttered was subtracted to compensate for ambient light and detector dark currents. The intensities of the background-subtracted spectra were calibrated relative to a spectrum that was acquired from a white reflectance standard before all of the insertions. For each location of the needle tip, the optical spinal needle provided a spectrum that derived from tissue located approximately 1.2 mm or closer to the bevel surface. Spectra were acquired with an integration time of 0.5 s.

Image guidance was provided with a 20-cm flat detector angiography system (Allura Xper FD20; Philips Healthcare, Best, The Netherlands), which allowed for both two-dimensional fluoroscopic imaging and three-dimensional rotational C-arm computed tomography (XperCT).11XperCT imaging was based on a motorized circular movement of the C-arm with a motion range of 240°, during which 620 rotational fluoroscopic frames were acquired. Reconstruction and visualization were performed with a three-dimensional rotational angiography workstation.

Before each insertion, the needle trajectory was planned in software with the three-dimensional rotational angiography workstation following XperCT imaging.12This software uses XperCT images to calculate the angle of insertion and distance required to advance a needle along a direct path from the skin entry point to the desired target, which in this study was the epidural space. At the location on the skin identified by the planning software, an introducer needle was first inserted superficially such that the tip reached a depth of approximately 5 mm relative to the skin surface. It was subsequently removed, and the optical spinal needle was inserted along this predetermined path. Previous placement of the rigid introducer needle facilitated insertion of the more flexible optical spinal needle through the thick swine skin.

Numerous image-guided needle insertions to the epidural space were performed. The needle trajectory from the skin's surface to the epidural space was altered to simulate various needle paths that are common during localization of the epidural space, including midline insertion, lateral deviation superficial to the epidural space, and a paramedian approach. Insertions were carried out at lumbar and thoracic levels. During one insertion, the needle entered an epidural vein on entry into the epidural space, allowing measurement of the spectral characteristics of this common problematic needle position. During each insertion, the needle was advanced through anteriorly beyond the epidural space to allow measurement of the spectral characteristics of entry into the cerebrospinal fluid (CSF), spinal cord, and the intervertebral disc. We note that the conus medullaris of the spinal cord typically extends to the sacral level in swine.13 

During each insertion, the optical spinal needle was advanced in discrete increments aimed at placing the needle at specific anatomic levels (e.g ., in the muscle superficial to the interspinous ligament, within the interspinous ligament, and within the epidural space). The needle was held in precise position at each depth using a custom-built radiolucent clamp. Anteroposterior and lateral fluoroscopic guidance was used to judge the approximate needle depth at each level, and the precise anatomic position was then determined using XperCT images. At each needle position, 20 optical spectra were acquired in addition to an XperCT image. Positioning of the needle within the epidural space was based on the XperCT images and was performed without use of the loss-of-resistance method to avoid the introduction of air or saline into the epidural space, which would distort the optical spectra. During the image-guided insertions, spectra were not displayed to the practitioner performing the insertions to assure that the identification of the epidural space was determined anatomically using the XperCT images and was not affected by the spectroscopic information.

After the image-guided insertions, five insertions were performed in which guidance to the epidural space target was performed solely with the interpretation of optical spectra. These optically guided insertions were performed at consecutive lumbar levels with midline and paramedian approaches. During the insertions, spectra were acquired continuously, with a tissue spectrum and a corresponding background spectrum acquired approximately once every second; they were displayed to an independent member of the research team who interpreted them in real time without knowledge of the needle position. Interpretation was performed by visually comparing spectra displayed during the procedure with spectra acquired during the image-guided insertions. Specifically, the interpreter sought to identify a prominent lipid absorption peak centered at 1,210 nm that was similar to those observed in spectra previously acquired from the epidural space with x-ray guidance. After the identification of this peak, the needle was fixed in position, and an XperCT image was acquired to determine the anatomic position of the needle tip. Spectral analysis, as described in the following section, was not performed in real time and was not available to the interpreter during this pilot study. The practitioner who performed the optically guided needle insertions was blinded to the optical spectra and did not have x-ray image guidance; the only feedback provided during a given insertion occurred at the end, with an alert from the interpreter indicating that the prominent lipid absorption peak had been identified.

To obtain quantitative estimates of chromophore volume fractions from diffuse reflectance spectra, a computational model of light propagation was used. The model accepted chromophore volume fractions and light-scattering parameters as input and generated a spectrum as output. Estimates were revised iteratively to minimize the difference between the spectrum generated by the model and acquired from tissue. This method of spectral analysis was first developed to analyze spectra acquired from colon polyps14and was recently modified to include absorption by lipids in a study by Nachabéet al .15It was assumed that the tissue regions from which the spectra was derived were optically homogeneous. The light propagation model was applied separately to the spectra obtained from the two spectrometers. For spectra in the range of 500–1,000 nm, the dominant chromophores were assumed to be oxyhemoglobin and deoxyhemoglobin. For spectra in the range of 900–1,600 nm, the dominant chromophores were assumed to be water and lipid. With the aforementioned iteration method, two parameters were estimated: the blood fraction and the lipid fraction. The former parameter is defined as the estimated total concentration of hemoglobin (the sum of the concentrations of oxyhemoglobin and deoxyhemoglobin) and is expressed as a percentage of the total concentration of hemoglobin in normal human blood (150 g/liter). The latter parameter is defined as the estimated volume fraction of lipid normalized by the sum of the estimated volume fractions of lipid and water. Analysis of each spectrum took 4.5 s, with processing performed in MATLAB (Mathworks, Natick, MA) on a 2.4-GHz laptop computer.

Two statistical analyses were performed to determine the significance of the changes in blood and lipid fractions obtained from image-guided insertions. The first analysis focused on changes in the parameters corresponding to transitions in individual needle insertions. For each pair of adjacent tissues, a two-tailed t  test for paired samples (unequal variances) was applied to compare the parameters obtained with the needle tip at one location with those obtained with the needle tip at the second location. To examine the repeatability of using the optical guidance technique, a second analysis was carried out by combining the blood and lipid fraction values from all five needle insertions. Two-tailed t  tests for paired samples (unequal variances) were applied to compare the mean blood and lipid fractions obtained with the needle tip (1) in muscle versus  in the epidural space and (2) at ligamentous tissues just superficial to the epidural space (interspinous ligament or the ligamentum flavum) versus  in the epidural space. With all analyses, those yielding a P  value <0.05 were considered significant. Bonferroni corrections were applied to adjust for multiple comparisons; the level of significance for each analysis was set as follows: insertions 1 and 3, 4 comparisons, P < 0.0125 (0.05/4) was considered significant; insertions 2 and 5, 3 comparisons, P < 0.0167 (0.05/3) was considered significant; insertion 4, 2 comparisons, P < 0.025 (0.05/2) was considered significant.

Sagittal and axial cross-sections extracted from the XperCT images provided excellent visualization of the needle, the spine, and the soft tissues, including the spinal cord and posterior skeletal muscles, allowing precise anatomic localization of the needle tip. During needle insertion in the midline at the lumbar level (fig. 1A), sagittal and axial XperCT images show the position of the needle tip at five different depths. The epidural space—the adipose-rich region surrounding the spinal cord—could clearly be identified within the vertebral column as a thin region of lower x-ray attenuation relative to the CSF-containing thecal sac. The needle boundaries were slightly blurred because of x-ray attenuation discontinuities; at the distal part of the needle, these artifacts were slightly more pronounced. With surface reconstructions of the XperCT image volumes, the needle trajectory and the spinal geometry were clearly visualized, as shown for the first midline insertion in figure 1B.

The corresponding optical spectra obtained at each needle position are displayed as the intensity of light received from tissue as a function of the wavelength of light (fig. 1A). Optical spectra varied as the needle tip was positioned in different tissues. These variations could be characterized in terms of absorption peaks, which manifested as lower intensities within specific wavelength ranges. There was good correspondence between the spectral locations of the observed absorption peaks and those of known chromophores. It is likely that absorption peaks present between 500 and 600 nm derived in large part from hemoglobin.16The predominance of deoxygenated or oxygenated forms could be inferred from the shape of the spectra between 500 and 600 nm, with the former having a single absorption peak at 557 nm and the latter with two at 542 and 576 nm. An additional small absorption peak centered at 757 nm was probably associated with the presence of the deoxygenated forms. The presence of lipid is probably responsible for absorption peaks centered at 930 nm and 1,210 nm,17,18with the latter more pronounced; the presence of water is responsible for absorption peaks observed at 976, 1,197, and 1,455 nm.19The blood and lipid fractions are shown together with the corresponding optical spectra in figure 1A; these same parameters are concatenated in figure 1Cto show their variation as the needle tip progressed ventrally, demonstrating their potential for tissue differentiation.

Figure 1illustrates a midline approach to the lumbar epidural space. The transition of the needle tip from skeletal muscle to the epidural space was associated with a large increase in the lipid fraction (mean ± SD: 5.3 ± 0.4% to 47.6 ± 0.4%) and a decrease in the blood fraction (2.06 ± 0.01% to 0.183 ± 0.002%). With the needle tip in the interspinous ligament, intermediate values were observed for the two parameters (23.30 ± 0.09% and 0.769 ± 0.006%, respectively). As the needle tip progressed beyond the epidural space through the intrathecal space and into the intervertebral disc, little change in the blood fraction was observed; the lipid fraction decreased to 1.9 ± 0.1%. Figure 2illustrates entry into a vein within the epidural space. The transition of the needle tip from skeletal muscle to the epidural space was associated with an increase in the lipid fraction (19.1 ± 0.6% to 53.2 ± 0.2%), which was similar in magnitude to that observed in the first midline insertion. With the needle tip at the interspinous ligament, an intermediate lipid fraction was observed again (28.7 ± 0.3%). With the needle tip in the epidural space, the mean blood fraction was more than 7-fold higher than those at the previous two locations and 27-fold higher than it was in the first midline insertion. When the stylet was removed with the cannula in place, spontaneous backflow of venous blood consistent with the puncture of an epidural vein was observed. With the needle tip in the intrathecal space, the mean blood fraction was higher than in the first midline insertion, and the mean lipid fraction was lower; removal of the stylet revealed backflow of blood-tinged CSF.

Figure 3illustrates deviation of the needle tip from the interspinous ligament to the lumbar paravertebral skeletal muscle, superficial to the epidural space and then through the lateral portion of the ligamentum flavum and into the epidural space. The transition from interspinous ligament to the paravertebral muscle was associated with a decrease in the lipid fraction (57.4 ± 0.3% to 38.2 ± 0.8%) and an increase in the blood fraction (0.596 ± 0.006% to 1.26 ± 0.01%). Likewise, with the first two insertions, transition of the needle tip from skeletal muscle to the epidural space resulted in a higher lipid fraction (74 ± 1%) and a lower mean blood fraction (0.241 ± 0.006%). With the needle tip at the ligamentum flavum, intermediate values were observed for the two parameters (25.1 ± 0.4% and 0.36 ± 0.02%, respectively). The transition of the needle tip from the epidural space into the intrathecal space was associated with a decrease in the lipid fraction and an increase in the blood fraction (fig. 3). With the needle tip in the intrathecal space, removal of the stylet revealed backflow of clear CSF.

Figure 4illustrates a paramedian approach to the lumbar epidural space, traversing from the paraspinous musculature, through the ligamentum flavum and in the epidural space near midline. Again, an increase in lipid fraction was observed with the transition of the needle tip from skeletal muscle to the epidural space (3.6 ± 0.2% to 73.2 ± 0.5%), and an intermediate value (38.9 ± 0.3%) was once again observed with the needle tip at the ligamentum flavum. As with the third insertion, a decrease in the blood fraction was also observed with the transition from skeletal muscle to ligamentum flavum in the fourth insertion; in contrast, the transition from ligamentum flavum to the epidural space transition was associated with a slight increase in the blood fraction.

Figure 5illustrates midline needle insertion at the thoracic level. Variations in the lipid fraction were similar to those in the first and second insertions (lumbar midline), with increases observed as the needle tip progressed from skeletal muscle (32.3 ± 0.2%) to the epidural space (69.4 ± 0.3%) via  the interspinous ligament (42.4 ± 0.3%). Interestingly, the blood fraction was high when the needle tip was directly within the spinal cord, but backflow of blood was not observed; at that location, the lipid fraction was lower than that when the needle tip was in the epidural space.

Across all epidural insertions, the mean lipid fractions obtained from skeletal muscle were 1.9- to 20-fold higher than those obtained from the epidural space. Transitions of the needle tip from adjacent ligamentous tissues (the interspinous ligament and the ligamentum flavum) to the epidural space were also associated with increases in the mean lipid fraction, which ranged from 1.6- to 3.0-fold.

The changes in both blood and lipid fractions corresponding to transitions from one tissue type to another for individual needle insertions were found to be significant in all cases (P < 0.001). For every transition, the pairwise comparisons that were tested for each needle insertion are shown graphically in figures 1C–5C; the differences in blood and lipid fractions are presented in table 1. The mean lipid fraction obtained with the optical needle tip in the epidural space was significantly higher than that obtained from adjacent ligamentous tissues for all five needle insertions (P = 0.0020). The mean lipid fraction obtained with the optical needle tip in the epidural space was also significantly higher than that in muscle for all five needle insertions (P = 0.0013). The corresponding differences in the mean blood fraction were not significant. Differences in the mean blood and lipid fractions obtained from all insertions are summarized in table 2.

In each insertion that was guided solely with real-time interpretation of optical spectra, the interpreter identified a prominent absorption peak centered at 1,210 nm. These absorption peaks were visually consistent with those identified in spectra acquired from the epidural space during x-ray guided insertions. Examination of the XperCT images that were acquired after these prominent absorption peaks had been reported revealed that the needle tip had reached the epidural space at the end of each of the five consecutive insertions.

The traditional LOR technique depends on unreliable and somewhat subjective measures of the mechanical resistance to injection of air or saline. In contrast, spectroscopic detection of the epidural space employs quantitative and reproducible parameters. The prominent lipid absorption peaks that were observed in spectra acquired from the epidural space are consistent with cryomicrotome studies, which found a predominance of homogeneous fat in the posterior epidural compartment.20This histologic correlate could prove to be advantageous in terms of interpreting the optical spectra. With parameters that are based on reflectance at visible wavelengths, such as those developed in a previous study,10histologic correlates remain to be determined; furthermore, one can expect that minor bleeding events could have a significant influence on their values. Optical identification of the epidural space with the lipid parameter may have limitations in certain clinical contexts, however, such as access to the epidural space in the cervical spine where there is a relative paucity of epidural fat21or with certain pathologic conditions such as spinal stenosis that are frequently accompanied by local reductions in the amount of epidural fat. The relatively small distance in front of the needle sampled for the spectral data (estimated to be approximately 1.2 mm) may well allow use, however, even when the dimensions of the epidural space are limited.

In one of the five insertions, epidural access was associated with a large increase in the blood fraction that coincided with venous backflow. That coincidence gives confidence that the optical spinal needle could provide a means for early detection of intravascular needle position, before local anesthetic or an epidural catheter are placed. However, the lipid fraction remained high, suggesting that the volume of tissue sampled at this intravenous location included both the blood around the needle tip and some of the adjacent epidural fat. More detailed studies are required to determine the distance from the bevel surface at which veins and arteries can be detected with optical spectroscopy. It is possible that perivascular needle positioning can be detected even before the vessel is entered, based on the spectral characteristics of the exterior wall of the vessel. Given the differences in the wavelength dependence of light absorption by oxygenated and deoxygenated forms of hemoglobin, refinements of the spectral processing algorithm should allow for the hemoglobin oxygenation saturation at the needle tip to be quantified. Accurate hemoglobin oxygenation saturation measurements would allow for differentiation of arterial and venous structures.

There were significant variations in both blood and lipid fractions when the needle tip was within the CSF, suggesting that the position of the needle tip relative to the thecal sac or neural elements may alter the spectral characteristics. However, transitions from the epidural space to the intrathecal space were consistently associated with large reductions of the lipid fraction. In all three insertions in which this transition was tested, the mean reductions ranged from 4.1- to 6.9-fold. This suggests that optical spectroscopy allows for ready detection of entry into the thecal sac, potentially obviating the need to use a test dose of local anesthetic as the only reliable test to rule out intrathecal placement. However, the measurement geometry is important for interpreting the spectroscopic information. Even when the opening of the needle cannula is completely within the intrathecal space, a part of the long bevel surface and the distal end of at least one optical fiber may be in adjacent tissues. Optical absorption in these adjacent tissues could produce a large lipid fraction, falsely leading the operator to conclude that the entire needle is within the epidural space. Another confounding factor is that the scattering of light in clear CSF is much lower than in tissues, so that measured light probably derived in part from tissues enclosing the intrathecal space.

Spectroscopic detection of spinal cord penetration could be valuable for preventing rare but potentially devastating injections within the spinal cord. The high blood fraction and the prominent deoxyhemoglobin absorption peaks observed after the needle tip penetrated the spinal cord (fig. 5) may have resulted from bleeding within the spinal cord after needle penetration. Further study is needed to determine to what extent optical spectroscopy can detect intracord needle tip placement.

Penetration into the intervertebral disc (fig. 1) is not typically a concern with fluoroscopically guided epidural steroid injections performed in clinical practice because this structure is so far anterior. The present study demonstrated a distinctive spectral pattern within the disc, another example of the potential for tissue differentiation with optical spectroscopy: both the blood and lipid fractions changed significantly during the transition of the needle tip from the epidural space to the intervertebral disc.

The optical signals acquired from a particular location in tissue probably depend to some extent on the tissues encountered at previous times during the insertion. For instance, as the needle tip progressed through tissues, a small volume of blood could have remained close to the bevel surface and altered the spectroscopic signals; the extent to which this effect manifested is unclear. Blood could also have diffused into the intrathecal space together with the needle tip, particularly if an epidural vein was punctured. This situation was likely encountered in the second insertion; it is consistent with the increased blood fraction observed when the needle tip was in the intrathecal space.

Several confounding factors can decrease the accuracy of the blood and lipid fraction estimates and their interpretation. For the blood fraction, they include the presence of additional tissue chromophores, such as myoglobin in muscle and carotenes in epidural fat.22The total hemoglobin content of blood in swine is broadly similar to that in humans, but variations as a function of age and sex are known to occur in both species23,24; as such, a degree of caution is warranted when interpreting relating blood fraction to whole blood. For the lipid fraction, confounding factors may include the presence of chromophores such as collagen.25Variations in optical scattering, such as those induced by tissue compression26during needle insertions, may affect the average distance traveled by light in tissue and thereby affect the volume of tissue from which a spectrum derives.

Variations in the angle of the needle bevel with respect to tissue layer interfaces probably contributed significantly to the intrainsertion differences of the lipid and blood fractions. This effect was identified in the study by Ting et al .10For instance, in cases where the needle tip was within muscle, spectroscopic signals could have been influenced by the presence of adjacent fascia, depending on the angle of the bevel. Likewise, in cases where the needle tip was in close proximity to the interspinous ligament, spectroscopic signals could have been influenced by the presence of adjacent adipose and skeletal muscle tissues. The pressure imparted by the needle on tissues may have altered the geometric arrangement of the tissue layer interfaces and the angle of the needle bevel. In cases where the needle tip was in contact with the ligamentum flavum and little pressure was applied, one could expect spectroscopic signals to derive primarily from skeletal muscle. When more pressure was applied, the ligamentum flavum surface may have been deformed, so that more light was primarily directed ventrally toward the epidural fat; one could expect a higher lipid fraction in this case.

Lateral deviation of the needle tip from the interspinous ligament to adjacent skeletal muscle during midline insertions occurs relatively frequently in clinical practice. This deviation can result in false LOR and failed epidural anesthesia when decreased resistance to injection of air or saline occurs as the needle tip enters muscle. Spectroscopic signals could differentiate between true LOR in the epidural space and false LOR in skeletal muscle. In this study, the latter event was associated with a lower lipid fraction and a higher blood fraction; the former, with a higher lipid fraction and frequently with a lower blood fraction.

Relative changes in the blood and lipid fractions that occur in a particular insertion may provide more robust indications of epidural space access than absolute values. For instance, the lipid fraction obtained from the epidural space in the first insertion was only 1.2-fold higher than that obtained from ligamentum flavum in the fourth insertion. However, in both of those insertions, epidural space access was associated with an increase in lipid fraction of at least 1.9-fold relative to those obtained from adjacent tissues. Relative changes in the blood and lipid fractions are not likely to be uniquely associated with specific tissue layer transitions, as hemoglobin and lipid concentrations vary widely in tissues encountered during epidural injections. Because the optical spectra are acquired from tissues that are very close to the bevel surface, we expect that correct interpretation of spectroscopic information may ultimately depend on image guidance to indicate the locations of anatomic landmarks relative to the needle tip (e.g ., fluoroscopy or ultrasound).

The encouraging results obtained in this in vivo  swine study set the stage for larger-scale studies. If similar spectroscopic changes associated with epidural access are observed consistently in humans, they could indicate the correct endpoints of epidural injections in real time. In this way, the optical spinal needle has the potential to increase the efficacy of epidural injections and reduce the incidence of complications, such as intravascular injection, dural puncture, intrathecal injection, and spinal cord damage. The optical spinal needle would therefore be complementary to fluoroscopic imaging, providing soft-tissue contrast for tissues that has never before been available.

The authors thank Pellina Janson, Technician, Karolinska Experimental and Research Imaging Center, Karolinska University Hospital, Stockholm, Sweden.