Ultrasonography of the Adult Thoracic and Lumbar Spine for Central Neuraxial Blockade

The first report of ultrasound-guided lumbar puncture appeared in the Russian literature in 1971.1Nine years later, Cork et al . described the use of ultrasound to delineate neuraxial anatomy.2Although the images were of poor quality by today's standards, they were able to define the lamina, ligamentum flavum, spinal canal, and the vertebral body. Thereafter, ultrasound was used mostly to preview the spinal anatomy and measure the distances to the lamina and epidural space before epidural puncture.3,4Between 2001 and 2004, Grau and colleagues conducted a series of investigations that demonstrated the utility of ultrasound in epidural analgesia and were pivotal in improving our understanding of spinal sonography.5–15Despite this, only three case reports appeared in the adult anesthetic literature between the end of 2004 and beginning of 2007,16–18and it is likely that the quality and availability of ultrasound imaging at the time hindered research in this area. Since then, there have been an increasing number of anesthesia-related publications (including a set of National Institute for Health and Clinical Excellence [NICE]) guidelines19) on ultrasound-guided epidural and spinal anesthesia. There also has been interest in the use of the technique by emergency physicians to guide lumbar puncture.20–23 

A typical vertebra has two components: the body and the arch. The vertebral arch is composed of the following elements: pedicles, lamina, transverse processes, spinous process, and superior and inferior articular processes (fig. 1). Adjacent vertebrae articulate at the facet joints between superior and inferior articular processes and at the intervertebral discs between vertebral bodies. In this article, we use the terms “interlaminar space” and “interspinous space” to refer to the gaps between adjacent laminae and spinous processes, respectively.

The vertebral canal is formed by the spinous process and lamina posteriorly, the pedicles laterally, and the vertebral body anteriorly. The posterior longitudinal ligament runs along the length of the anterior wall of the vertebral canal. The only openings into the vertebral canal are the intervertebral foramina along its lateral wall, from whence the spinal nerve roots emerge, and the interlaminar spaces on its posterior wall. The ligamentum flavum is a dense connective tissue ligament that bridges the interlaminar spaces. It is arch-like in cross-section and is thickest in the midline. The ligamentum flavum attaches to the anterior surface of the lamina above but splits to attach to both the posterior surface (superficial component) and anterior surface (deep component) of the lamina below.24The spinous processes are connected at their tips by the supraspinous ligament, which is a strong fibrous cord, and along their length by the interspinous ligament, which is thin and membranous.

Within the vertebral canal lie the thecal sac (formed by the dura mater and arachnoid mater) and its contents (spinal cord, cauda equina, and cerebrospinal fluid). The epidural space is the space within the vertebral canal but outside the thecal sac. The anatomy of the epidural space is more complex than is portrayed in most anatomy textbooks.25It is divided into anterior, lateral, and posterior epidural spaces with respect to the thecal sac, with the posterior epidural space being of most interest in neuraxial blockade. The posterior epidural space is not continuous. Instead, it is segmented into a series of fat-filled compartments in the interlaminar areas. The lateral epidural spaces are located at the level of each intervertebral foramen and contain spinal nerves, radicular vessels, and fat. The primary structure of importance in the anterior epidural space is the internal vertebral venous plexus.

The posterior surface of the laminae of the five lumbar vertebrae slopes in an anterosuperior direction (fig. 1). The laminae, unlike in the thoracic spine, do not overlap, and there is a distinct interlaminar space between adjacent vertebrae. The spinous processes are broad and flat in the vertical dimension and project posteriorly, with only a slight inferior angulation. Thus, accessing the vertebral canal in the midline via  the interspinous and interlaminar spaces is relatively easy. These spaces are further enlarged by forward flexion.26Midline access can be more difficult in the elderly because of narrowing or calcification of the interspinous space, heterotopic ossification of the interspinous ligaments,27and hypertrophy of the facet joint. The transverse processes arise anterior to the articular processes and project posterolaterally; the L3 transverse process is characteristically the longest.24The facet joints and transverse processes lie in approximately the same transverse plane as the interlaminar space, and the inferior edge of the spinous process overlies the widest part of the interlaminar space.

The ligamentum flavum arches over the interlaminar space; deep to it lies the fat-filled compartment of the posterior epidural space (fig. 2). The posterior epidural space has a triangular cross-section (typically 7 mm wide in the midline anteroposterior dimension) in the lumbar region and narrows away to a virtual space anterior to the laminae, where the posterior dura lies in direct contact with bone.25Within the thecal sac, the conus medullaris in the adult is most often located at the level of the first lumbar (L1) vertebral body; however, its location in any individual patient follows a normal distribution and may range from the middle of the twelfth thoracic (T12) vertebra to the upper third of L3.28The conus medullaris gives rise to the cauda equina and filum terminale. The thecal sac typically ends at the midpoint of the second sacral vertebra (S2), although in the individual patient this can range from the upper border of S1 to the lower border of S4.29 

During scanning of the lumbar spine, patients should be placed in the position in which the block is to be performed; this is usually the lateral decubitus or sitting position. We recommend a curved-array, low-frequency (2–5 MHz) probe because the wide field of view and deeper penetration improve recognition of anatomy and image quality, respectively. An initial depth setting of 7–8 cm is appropriate for most patients, but the depth, focus, and gain settings of the ultrasound machine should be adjusted as needed during the scanning process to produce an optimal image.

Human anatomy is characteristically described in terms of three basic planes: sagittal, transverse, and coronal (fig. 3). Similarly, there are three basic orientations of the ultrasound probe and beam: (1) paramedian sagittal (PS), when the beam is oriented in the sagittal plane of the spine lateral to the median (midline) sagittal plane; (2) paramedian sagittal oblique (PS oblique), similar to the PS plane except that the beam is now tilted and aimed toward the median sagittal plane; and (3) transverse, when the beam is orientated parallel to the transverse or horizontal plane. The terms “transverse” and “axial” are synonymous when referring to imaging planes; we shall be using the former term throughout this review.

Pattern recognition is essential in interpreting spinal sonoanatomy because the depth and limited acoustic windows often preclude clear visualization of the relevant anatomic structures. It is worth remembering that bony surfaces appear as hyperechoic (white) linear structures with dense acoustic shadowing (black) beneath that completely obscures any deeper structures. Connective tissue structures, such as ligaments and fascial membranes, also are hyperechoic; however, their acoustic impedance is less than that of bone, so deeper structures can still be imaged. Fat and fluid have very low acoustic impedance and are hypoechoic (dark). A systematic approach to scanning (table 1) facilitates both the process of pattern recognition and the overall performance of ultrasound-guided neuraxial blockade. There are five basic ultrasonographic views that may be obtained, and these are described here in detail.

To start, the ultrasound probe is placed in a PS orientation 3–4 cm lateral to the midline and just above the upper border of sacrum. In this view, the transverse processes of successive lumbar vertebrae are visualized. These appear as short hyperechoic curvilinear structures with pronounced “finger-like” acoustic shadowing beneath, an appearance that has been described as the “trident sign.”30The striated psoas major muscle is visible between the acoustic shadows and deep to the transverse processes (fig. 4).

From the PS transverse process view, the probe is slid medially until a continuous hyperechoic line of “humps” is seen (fig. 5). In this PS articular process view, each hump represents the facet joint between a superior and inferior articular process of successive vertebrae. Both the superior and inferior articular processes lie in the coronal plane posterior to the transverse processes and thus are seen at a more superficial depth than are the transverse processes.

Once the PS articular process view has been obtained, the probe is tilted to angle the beam in a lateral-to-medial direction toward the median sagittal plane. The sloping hyperechoic laminae of the lumbar vertebrae form a “sawtooth”-like pattern in this view. The intervening gaps represent the paramedian interlaminar spaces, through which the following structures may be visualized (in order, from superficial to deep): ligamentum flavum, epidural space, posterior dura mater, intrathecal space, anterior dura, posterior longitudinal ligament, and posterior vertebral body (fig. 6).

The ligamentum flavum, epidural space, and posterior dura often appear as a single linear hyperechoic structure, which we have termed the posterior complex. Small sliding and tilting movements of the probe may allow the ligamentum flavum and posterior dura to be distinguished as two hyperechoic lines separated by the hypoechoic fat-filled posterior epidural space. However, the posterior epidural space may not always be visible. This is partly explained by the limitations of ultrasound resolution, particularly in obese patients, but also by the posterior epidural space being triangular in cross section.25It thins significantly toward its lateral margins, so its apparent width depends on exactly where the ultrasound beam intersects it. The intrathecal space is uniformly hypoechoic, although the cauda equina and filum terminale may be visible as hyperechoic pulsatile streaks within the space. The anterior dura, posterior longitudinal ligament, and posterior aspect of the vertebral body or the intervertebral disc are collectively visible as a single linear hyperechoic structure (the anterior complex31) and are almost never distinguishable from one another in adults.

The superior-inferior dimensions of the interlaminar space may be estimated from the length of the posterior or anterior complex and may provide an indication of the technical difficulty associated with central neuraxial blockade at that level.32The depth from skin to the posterior complex may be measured to provide an indication of the expected needle depth for spinal or epidural anesthesia.33 

While the PS oblique view is maintained, the probe is slid in a caudad direction until the horizontal hyperechoic line of the sacrum comes into view (fig. 7). The gap between the line of the sacrum and the sawtooth of the L5 lamina is the L5–S1 intervertebral space. A characteristic of the L5 lamina is that it is narrower than the other lumbar laminae, and this may facilitate identification. The other lumbar interspaces are readily identified in the PS oblique view by counting upward from the lumbosacral junction. The surface location of each interspace may be indicated by centering it on the ultrasound screen and making a corresponding mark on the skin at the midpoint of the long edge of the probe (fig. 8). This prevents misidentification of the level during later scanning in the transverse plane.

Once the examination in the PS plane is completed, the probe is rotated 90 degrees into a transverse orientation and centered on the neuraxial midline. If the probe lies over a spinous process, the tip of the spinous process is visible as a superficial hyperechoic line with acoustic shadowing beneath. Its position may be marked, if desired, by centering it on the ultrasound screen as described above. The hyperechoic lamina is visible on either side of the spinous process, but all other structures of interest are obscured by bony acoustic shadowing (fig. 9).

Sliding the probe in a cephalad or caudad direction from the transverse spinous process view aligns the beam with the interspinous and interlaminar space and provides a transverse interlaminar view of the contents of the vertebral canal. Typically, the linear acoustic shadow of the spinous process gives way to a less dark vertical line (the interspinous ligament framed by the adjacent echogenic erector spinae muscles) and, deep to this, the two parallel hyperechoic lines of the posterior and anterior complex separated by the hypoechoic intrathecal space (fig. 10). Depending on the width of the interspinous space and the angle at which the spinous processes project, the transducer may have to be tilted cephalad to optimize the image of the vertebral canal.

Unlike in the PS oblique view, in the transverse interlaminar view the ligamentum flavum and posterior dura are rarely visible as distinct structures14,33,34and occasionally may not be visible.34The poorer view of the posterior complex may be attributed to the narrower acoustic window that exists between spinous processes; however, it has also been suggested that absence of the posterior complex is caused by physical gaps in the ligamentum flavum.34,35If the anterior complex is visible, the beam has traversed the vertebral canal, and one can be confident that the interlaminar space has been identified. The transverse processes and articular processes are additional helpful landmarks in difficult cases because they lie in approximately the same transverse plane as the interlaminar space.

Once an optimal view has been obtained, the depth from skin surface to the posterior complex may be measured using the electronic caliper built into the ultrasound machine. The neuraxial midline and the interlaminar space correspond with the midpoint of the long and short sides of the probe, respectively and can be marked on the skin (fig. 8). The intersection of these two landmarks indicates a suitable needle insertion point for a midline approach to spinal or epidural anesthesia. The cephalad angulation required to enter the interlaminar space also can be estimated from the degree of probe tilt required to obtain an optimal transverse interlaminar view.

The morphology of the 12 thoracic vertebrae varies throughout the length of the thoracic spine. The first four thoracic vertebrae (T1–T4) are similar to the cervical vertebrae in some respects; they have vertically oriented articular processes and spinous processes that project directly posteriorly. The lowermost four vertebrae (T9–T12) are similar to the lumbar vertebrae; their articular processes project laterally, and their spinous processes are broad, flat, and project directly posteriorly. On the other hand, the spinous processes of T5–T8 vertebrae project posteriorly at an extreme inferior angle, such that the inferior border of the spinous process overlies the midpoint of the lamina of the vertebra below (fig. 11). The laminae of adjacent thoracic vertebrae are also overlapping, making the interlaminar spaces in the thoracic spine extremely small and difficult to access. The thoracic transverse processes arise posterior to the articular processes and articulate with the corresponding rib. The presence of a rib is an identifying feature of the transition between L1 and T12 vertebra and can be used in conjunction with the “counting-up” approach from the L5–S1 junction to determine the intervertebral level.

The ultrasonographic appearances of the lower thoracic (T9–T12) vertebrae and the lumbar vertebrae are similar, except that the interlaminar spaces tend to be narrower (fig. 11). With the transducer in a PS orientation 2–3 cm lateral to the midline, the ribs are visible as short hyperechoic lines with pronounced acoustic shadowing beneath. Sliding the transducer medially produces a PS articular process view similar to that of the lumbar spine. The PS oblique and transverse views are then obtained as described in the section on lumbar spine imaging.

Imaging in the midthoracic spine is much more difficult because of the extreme caudad angulation of the spinous processes and the overlapping laminae. In practice, we find that although the spinous process, lamina, transverse processes, ribs, and pleura are visible on scanning in the transverse plane, it is nearly impossible to obtain a transverse interlaminar view (fig. 12). Thus, the transverse scan provides very little information relevant to neuraxial blockade apart from identifying the midline and measuring the depth to the lamina. On the other hand, the PS oblique view is more useful. Here, the laminae are visible as horizontal hyperechoic curvilinear structures with acoustic shadowing beneath, and although the narrow width of the interlaminar space may prevent visualization of the intrathecal space and anterior complex, the location of the interlaminar spaces can be readily identified and marked in the same manner as for the lumbar region (fig. 13).

We performed a literature search for relevant studies in the MEDLINE database for the period from its inception until October 22, 2010. We limited search results to human studies in adults (≥19 yr). The electronic search strategy contained the following MeSH and free-text terms: (spine OR spinal OR epidural OR neuraxial OR caudal) AND (ultrasound OR ultrasonography OR ultrasonographic) AND (anesthesia OR analgesia OR block). This yielded 875 articles. We reviewed the title, abstract, and as appropriate, the full text of these articles. The reference lists of the selected articles and the authors' personal file collections also were consulted to identify any studies missed by the electronic search strategy.

This resulted in a list of 55 relevant articles. The breakdown by study type is as follows: 7 review articles,6,31,36–405 randomized controlled trials (RCTs),8–10,12,4127 observational cohort studies,2–5,7,13,14,16,33–35,42–5714 case reports,17,18,32,58–68and 2 technical articles.69,70Most (62%) of the clinical reports involved obstetric patients. Only three articles pertained to ultrasound imaging of the thoracic spine.11,16,63The methodology and results of the RCTs and observational studies are summarized in tables 2 and 3.

Four RCTs compared the ultrasound-guided technique to the conventional surface landmark-guided technique and examined outcomes related to the clinical efficacy of neuraxial blockade.8,10,12,41All involved obstetric patients receiving epidural or combined spinal-epidural anesthesia. In three of these studies,8,10,12interventions and outcomes were performed and assessed by the same (unblinded) investigator; thus, caution is warranted in extrapolating the results. In the largest of these studies10(n = 300), a significantly lower rate of incomplete analgesia (2 vs . 8%, P < 0.03), as well as lower postblock pain scores (scale 0–10, 0.8 ± 1.5 vs . 1.3 ± 2.2, P = 0.006) were seen in the ultrasound-guided group. Patient satisfaction scores were significantly higher in two of the studies,8,10although the differences do not appear to have been clinically important (table 2). Nonsignificant trends to a lower rate of asymmetric block and patchy block were seen in all three studies.

More recently, Vallejo et al .41randomized 370 parturients receiving labor epidurals into two groups. One group underwent preprocedural ultrasound imaging of the lumbar spine by a single operator with 6 months' experience in the technique, and the other group did not. All epidurals were performed by a cohort of 15 first-year anesthesiology residents. The information obtained from the ultrasound scan (depth to the epidural space and location of landmarks) was communicated to the resident performing the epidural, who was subsequently supervised by another blinded staff anesthesiologist. The epidural failure rate (defined as inadequate analgesia requiring replacement of the epidural) was significantly lower in the ultrasound-guided group of patients (1.6 vs . 5.5%, P < 0.02).

Thus, evidence suggests that the ultrasound-guided technique improves the success and quality of epidural analgesia. However, most of the data originate from a single investigator, and additional randomized trials are needed to establish whether this benefit can be realized by less-experienced practitioners.

The technical difficulty of neuraxial blockade may be measured using two parameters: the number of needle manipulations required for success and the time taken to perform the block. Of the two, we consider the former to be more important because multiple needle manipulations or passes are an independent predictor of complications, such as inadvertent dural puncture, vascular puncture, and paresthesia.71In turn, elicitation of paresthesia is a significant risk factor for persistent neurologic deficit after spinal anesthesia.72–74Data from the five RCTs indicate that use of the ultrasound-guided technique either halved the number of needle passes required for successful neuraxial blockade8,10,12,41or significantly increased the first-pass success rate (75 vs . 20%, P < 0.001).9In another comparative nonrandomized trial, the success rate of residents learning to perform labor epidurals was significantly increased and accelerated by providing them with information obtained from a preprocedural ultrasound scan.5Again, it should be noted that five of these six studies were conducted by the same investigator and are thus susceptible to bias.

It is only logical that ultrasound would be most helpful in patients with poor or abnormal anatomic landmarks, and this is supported by numerous case reports of successful ultrasound-guided neuraxial block in patients with marked obesity (five reports),17,20,62,67,75previous spinal surgery and instrumentation (seven reports),18,59–61,65,66,68and spinal deformity (four reports).16,32,58,63In one of the five published RCTs, Grau et al .8specifically enrolled 72 parturients in whom neuraxial block was anticipated to be difficult because of the presence of spinal deformity, obesity (body mass index more than 33 kg/m²), or a history of previous difficulty. Patients in whom ultrasound imaging was used underwent fewer needle passes (1.5 ± 0.9 vs . 2.6 ± 1.4, P < 0.001) at fewer spinal interspaces (1.3 ± 0.5 vs . 1.6 ± 0.7, P < 0.05) than did the control group.

The lead author of the present paper recently completed a RCT of ultrasound-guided spinal anesthesia in 120 patients with difficult anatomical landmarks (defined as the presence of poorly palpable surface landmarks and a body mass index >35 kgm^-2, significant spinal deformity, or spinal surgery resulting in distortion or absence of surface landmarks).⁷⁶This study involved multiple experienced operators, each of whom performed both landmark identification (by palpation or ultrasound) and the spinal anesthetic itself. The primary outcome was the success rate of dural puncture on the first needle insertion attempt (this included needle redirections that did not involve complete withdrawal of the needle from the skin). There was a two-fold difference between the ultrasound-guided group and the control group in the first-attempt success rate (62% vs 32%, P < 0.001), and the median number of needle passes required for success (6 vs 13, P = 0.003).76 

In summary, ultrasound imaging of the spine by an experienced operator increases the ease of performance of neuraxial block, particularly in patients in whom difficulty is anticipated. Ultrasound may also be able to predict the ease of performance of neuraxial block and thus influence clinical decision-making32; however, this has yet to be systematically investigated.

Knowledge of the depth from skin to the epidural or intrathecal space allows selection of a needle of appropriate length and may help prevent inadvertent dural puncture. The correlation between ultrasound-measured depth and actual needle insertion depth has been evaluated in multiple studies: 10 in obstetric patients2–4,7–9,41,43,44,53and 3 in nonobstetric patients.23,33,45Correlation was excellent in all studies (Pearson correlation coefficients, 0.80–0.99), whether measurements were made in the sagittal, PS oblique, or transverse views. Of six studies that analyzed the difference between the two depths, the ultrasound-measured depth tended to underestimate actual needle depth in four3,33,44,53and overestimate it in the other two.9,43The 95% confidence limits for the difference ranged from 5 to 15 mm (table 3). Suggested reasons for the discrepancy include differing trajectories of ultrasound beam and needle and tissue compression by the probe during ultrasound scanning (which may cause as much as a 5-mm change in depth52) or by the Tuohy needle during insertion.

Incorrect identification of the lumbar intervertebral level has been implicated in conus medullaris injury after dural puncture.77,78Although the spinal cord and surrounding cerebrospinal fluid have a similar hypoechoic appearance on ultrasound, the cord and conus medullaris can be identified in the young pediatric population because the outer surface and central canal of the spinal cord are visible as bright hyperechoic lines.79,80These details are not visible in adults because of the greater depth and narrower acoustic windows into the spinal canal, and currently the conus medullaris cannot be localized on ultrasound in adults.

However, ultrasound can identify the intervertebral levels by counting spinous processes or laminae upward from the sacrum; this method is more accurate than clinical estimation using the intercristal line.57In fact, agreement between clinical and ultrasonographic methods of identifying intervertebral levels has been observed to occur in only 36–55% of cases.48,50,55Both Whitty et al .55and Schlotterbeck et al .50found that when there was disagreement, the clinically determined level was usually lower than that determined by ultrasound. However, Locks et al .48observed that the clinically determined level was higher, rather than lower. Their finding may be explained by their basing clinical identification on the premise that the intercristal line corresponded to the L4–L5 interspace, but a separate study found that the L3–L4 interspace (as identified on ultrasound) corresponded to the intercristal line in most subjects.49 

However, ultrasound is not infallible. Compared with other imaging modalities, such as magnetic resonance imaging,54computed tomography,46and plain radiographs57of the lumbar spine, ultrasound accurately identified a spinous process or intervertebral space only 68–76% of the time. It is worth noting that any inaccuracy observed with ultrasound is likely to be within one interspace of the true level, rather than two or three interspaces, as may occur with palpation of surface landmarks. In addition, two of these three studies used ultrasound technology that would now be considered obsolete,54,57so this may have contributed to misidentification. Errors are also more likely in the early stages of learning to perform ultrasonography of the spine,46,56and accuracy rates of 90% or greater probably can be achieved with adequate training and experience.46Errors usually result from misidentification of the L5–S1 junction56or failure to recognize developmental anomalies of the lumbosacral junction, which occur in approximately 12% of the general population.81Sacralization of the L5 vertebra is most common, in which there is a degree of fusion between L5 and the sacrum involving one or both transverse processes. Less commonly, the S1 vertebra may resemble a lumbar vertebra (lumbarization). Complete sacralization or lumbarization that results in the presence of four or six true lumbar vertebrae, respectively, is a rare occurrence. Definitive diagnosis of lumbosacral transitional vertebrae requires plain radiographs of the spine,81which are not always available. However, the accuracy of ultrasound can be enhanced by combining a counting-up approach from the L5–S1 junction with a “counting-down” approach from the T12 vertebra (identified by the presence of the twelfth rib). Although an L1 accessory rib can be present in as much as 2% of the population,82the simultaneous presence of both anomalies is exceedingly rare. Finally, it is reassuring to note that Kim et al .83found the distance between the conus medullaris and Tuffier's line to be identical in patients with and those without lumbosacral transitional anomalies. Thus, they concluded it is clinically appropriate to count up from the apparent lumbosacral junction when choosing an appropriate level for administration of spinal anesthesia.

When pertinent structures such as the ligamentum flavum, dura mater, and anterior complex can be visualized in the thoracic spine, it is logical that ultrasonography should have the same utility that it does in lumbar neuraxial blockade. Currently, little has been published about this topic. Grau et al . performed an imaging study in 20 volunteers in which they demonstrated it was feasible to identify the pertinent anatomic landmarks with ultrasound imaging.11However, the authors noted that visualization of the epidural space was much more difficult than that of the lumbar spine, and the PS oblique view was the best for this purpose. The principal limitations of this small study are that only young, slim patients with normal spinal anatomy were included and only the T5–T6 interspace was studied.

As with lumbar neuraxial blockade, the main advantage of the ultrasound-guided technique may be in the patient with abnormal spinal anatomy. The use of ultrasound to delineate spinal anatomy before insertion of an epidural catheter in patients with scoliosis has been described in a single case report and a small case series. Pandin et al . used ultrasound to identify a suitable interlaminar window and measure the depth to the epidural space before inserting a midthoracic epidural catheter.63Accurate placement of the catheter was further confirmed by electric stimulation through the epidural needle and catheter. McLeod et al . used ultrasound to measure the degree of axial rotation in the thoracic spine.16This was done by placing the transducer in a transverse orientation between spinous processes and manipulating it until the hyperechoic laminae on either side of the midline were level on the ultrasound screen; rotation was then measured as the angle between the long axis of the transducer and the patient's sagittal plane. The least-rotated interspace was identified and used for epidural insertion via  a midline approach. Epidural insertion was successful at the chosen interspace in 8 of 11 patients and at the interspace above in the remaining 3. It is notable that a fairly basic ultrasound machine and a linear-array transducer were used in both reports.

In our opinion, even if the vertebral canal is not clearly visible, a preprocedural scan may provide information that will facilitate thoracic epidural catheter insertion. Apart from determining axial rotation (as described by McLeod et al .16), the depth to the lamina may be measured (as a surrogate marker of depth to the epidural space), the levels of the thoracic interspaces may be determined more accurately, and the locations of the midline and interlaminar spaces can be marked on the skin. Triangulation using this information will facilitate estimation of the appropriate needle insertion site and trajectory for a paramedian or midline approach. Currently, no published data support or refute these assertions.

Visualization of the deeper structures in the vertebral canal (epidural space, dura, intrathecal space, and anterior complex) can be difficult in certain patient populations.

In obese patients, structures are often less distinct because of the attenuation that occurs as ultrasound waves travel a greater distance through soft tissue. A phase aberration effect caused by the varying speed of sound in the irregularly shaped adipose layers also has been described.84However, advances in imaging technology (e.g ., compound imaging and tissue harmonic imaging) can compensate for this deterioration in image quality, and recent studies support the feasibility of ultrasonography in the obese population.33,44,76Simple measures should not be neglected, such as reducing the beam frequency to provide better penetration, adjusting the focus to the appropriate depth, and applying adequate pressure to improve skin-transducer contact and compress the overlying soft tissue. At a minimum, the spinous processes (indicating the midline) and interspinous gaps usually can be identified.67Successful entry into the interlaminar space is more likely if needle redirections from the initial insertion point are made in very small increments. The use of a 22-gauge or larger needle, particularly at lengths of more than 90 mm, should be considered because such needles are less likely to be deflected from their intended trajectory during insertion.

The problem in elderly patients is narrowing of the interspinous spaces and interlaminar spaces caused by ossification of the interspinous ligaments and hypertrophy of the facet joints, respectively.27Prominent spinous processes in a thin patient also can hinder adequate skin-probe contact and contribute to poor visualization. In such patients, obtaining a transverse view of the vertebral canal may be physically difficult or impossible, and the PS oblique view may be a better choice. Contact may also be improved by using a probe with a smaller footprint.

There is an inherent degree of inaccuracy when marking the needle insertion point on the skin during the preprocedural scan. Currently available curved-array probes do not have markings that precisely indicate from where the ultrasound beam emanates. There is also an element of tissue distortion when performing the ultrasound scan, particularly in the elderly, who often have loose and mobile skin. Finally, skin marking does not indicate the caudad-to-cephalad angle at which the needle must be advanced in a midline approach. This can be estimated only from the angulation of the probe required to produce an optimal image of the interlaminar space. However, these factors can be compensated for by experience with the ultrasound-guided technique.

As a result of these limitations, extensive experience with the ultrasound-guided technique may be required before competence is attained. In virtually all published studies to date, ultrasound imaging has been performed by a small number of experienced investigators.

Two small studies attempted to examine the learning curve associated with ultrasound imaging of the lumbar spine. Margarido et al .56recruited 18 anesthesiologists with no previous experience in ultrasound imaging of the spine and provided them with comprehensive training that included reading material, an educational video, a 45-min lecture, and a 30-min hands-on workshop. The subjects were assessed 7–14 days later on their ability to perform three tasks in a human volunteer with normal (“easy”) anatomy: identify lumbar intervertebral spaces, mark an optimal insertion point, and measure the depth to the epidural space. Accuracy was determined by comparing their performance with that of three experts. Each subject performed as many as 20 consecutive trials, and cusum analysis was used to determine whether competence was achieved. Only five (27%) subjects achieved competence in identifying the intervertebral spaces; none demonstrated competence at the other two tasks. However, these results are inconclusive because only 11 (61%) of the subjects managed to complete 20 trials in the allotted time of 1 h. The criteria for success were also very strict, and the authors noted that most of the errors did not stem from an inability to recognize the relevant anatomy, but rather from imprecision in skin marking and depth measurement. They concluded that these errors could have been avoided by greater meticulousness on the part of the operator.

Halpern et al .46also used cusum analysis to determine the learning curve associated with using ultrasound to identify a given spinous process accurately (subsequently confirmed by computed tomography). They studied two anesthesiologists with no previous experience in ultrasound imaging of the lumbar spine who received training on five patients each. Competence (defined as ≥ 90% accuracy) was achieved by one subject after examination of 22 patients; the other subject required examination of 36 patients before achieving competence.

These preliminary studies suggest that once basic knowledge on ultrasonography of the lumbar spine has been acquired, experience with 40 or more cases may be required to attain competence in scanning. This needs to be confirmed by larger and more robust studies. Additional research is also needed to determine the learning curve associated with the actual performance of a successful ultrasound-guided neuraxial block and optimal training strategies. Novel spine phantom models have been described that permit scanning and needle insertion to be practiced in a workshop setting; however, no data exist to demonstrate how effective these models are at knowledge and skills translation.69,70 

Most studies of ultrasound-guided neuraxial blockade have used preprocedural ultrasound imaging. There are only four published reports of lumbar central neuraxial blockade using continuous real-time ultrasound guidance. Grau et al .12used a two-operator technique; one operator manipulated the transducer in a PS oblique view while the other operator inserted the needle using a midline approach. Karmakar et al .47(epidural) and Chin et al .58(spinal) reported a single-operator technique in which a PS oblique view of the vertebral canal was obtained and the needle inserted in-plane with the ultrasound beam. In our opinion, the real-time ultrasound-guided approach is demanding technically, and more data are required before it can be recommended for routine use. There is also a risk of introducing ultrasound gel into the epidural or intrathecal space, the safety implications of which are unclear. Strategies to prevent this include using gel sparingly (e.g ., applied in a thin layer directly onto the probe surface, rather than the patient's skin) and ensuring that the needle insertion site is completely free of gel before puncture or using normal saline instead of gel as the coupling medium. More recently, an experimental technique using an on-screen overlay and fixed-needle guide has been described, which may reduce the difficulty associated with the freehand technique.52 

Ultrasound-guided neuraxial blockade is a useful technique that can, among other things, help practitioners more accurately identify intervertebral levels, estimate depth to the epidural space, and locate an appropriate interlaminar space for needle insertion. It is relatively easy to perform using the described systematic approach (table 1), but as with all new techniques, adequate training and clinical experience are required to realize its full potential. At this time, we do not believe the technique should supplant the traditional surface landmark-based techniques of spinal and epidural anesthesia; these are simple, safe, and effective in most patients. Instead, the utility of the ultrasound-guided approach is most evident in patients in whom technical difficulty is expected because of poor surface anatomic landmarks (e.g ., in obesity or after spinal surgery) or distorted spinal anatomy (e.g ., scoliosis).

The authors gratefully acknowledge the invaluable assistance of Cyrus C. H. Tse, B.Sc., Research Assistant, Department of Anesthesia, Toronto Western Hospital, Toronto, Ontario, Canada, in preparing the figures for this manuscript.