Chapter 8. Radiologic Evaluation of Spinal Disease

As the millennium has just passed, it is appropriate to review the significant advances in spinal imaging that have occurred in the preceding quarter century. Before then, plain film radiography, conventional tomography, and myelography with either gas or oily material as contrast agents had been the only methods available for imaging abnormalities involving the vertebrae, intervertebral disks, spinal cord, or cauda equina.

By 1975, a nonionic intrathecal contrast agent, metrizamide, was approved for clinical use. Unlike oily agents, nonionic contrast carried negligible risk for arachnoiditis, was absorbable, and thus eliminated the need for its removal from the thecal sac. Secondly, its neurotoxicity was minimal, compared with ionic water-soluble media, which never achieved widespread acceptance in the United States.

In 1977, the introduction of whole-body computed tomography (CT) permitted direct cross-sectional imaging of both spinal and paraspinal structures. However, the margins of the spinal cord could only be reliably demonstrated after the intrathecal administration of water-soluble contrast. This procedure is known as CT myelography (CTM). Because of the greater contrast sensitivity of CT, as compared with plain film myelographic technique, a smaller, less potentially neurotoxic dose of contrast agent could be administered for CTM. Nevertheless, this procedure still requires a lumbar puncture, with its attendant hazards to the patient.

By 1982, magnetic resonance imaging (MRI) became clinically feasible. MRI has proven to be superior to CT because the spinal cord and nerve roots could be visualized directly without the requirement for intrathecal contrast material. Most significantly, the parenchyma of the spinal cord could now be imaged and assessed for intrinsic pathology, such as multiple sclerosis plaques. These lesions may not alter the shape of the spinal cord, and, therefore, would be undetectable by CTM. Secondly, MRI provides multiplanar imaging, including sagittal and coronal orientations, with spatial and contrast resolution equivalent to the axial plane. Lastly, MRI poses no known health risk as it uses only radiofrequency energy, not ionizing radiation as is the case with CT.

Sagittal plane MRI is ideal for extended, rapid evaluation of the entire vertebral column, a procedure facilitated because the spine is arranged in a sagittal plane. Recent improvements in MRI receiver coil design (phased array coil) have provided the capacity to image the entire spine with excellent detail in less than 10 minutes (Fig. 8-1). This is especially helpful in the evaluation for spinal metastases, as these patients are often in pain, and thus have difficulty in remaining motionless for MRI. Newer pulse sequences, including half-Fourier turbo-spin echo (HASTE), can provide interpretable scans in less than 10 seconds, albeit with reduced spatial resolution compared with conventional magnetic resonance studies. HASTE imaging can also suppress some metal-induced artifacts arising from surgical hardware (Fig. 8-2), allowing improved visibility of anatomy or pathology otherwise obscured by these artifacts.

Although CT is capable of producing sagittal plane spinal scans, these can only be obtained by reformatting the data from prior axial plane scans, as positioning the patient for direct sagittal spine scanning within the gantry aperture is not physically feasible. Moreover, the number and thickness of the axial scans limits the longitudinal coverage and spatial resolution of the reformatted sagittal CT images (Fig. 8-3). Such restrictions do not apply to sagittal plane MRI, which acquires data in any selected plane with equivalent, selectable spatial and contrast resolution. Moreover, MRI can image pathology not apparent on CT. Intrinsic spinal cord disease, including syringomyelia or multiple sclerosis plaques, is uniquely observable by MRI. MRI is performed by subjecting the patient to a sequence of radiofrequency (RF) pulses while he or she is positioned within a strong, fixed magnetic field. By the use of various combinations of these RF pulses, known as pulse sequences, images can be created that are, in effect, a map of the differing energy-releasing characteristics of tissues, known as the T1 and T2 relaxation constants, within a selected plane of anatomy. In general, scans emphasizing T1 characteristics (i.e. T1 “weighted”) clearly define the outline of an anatomic structure, such as the spinal cord, whereas so-called T2-weighted images are more sensitive in distinguishing a region of pathology, including edema surrounding a tumor, gliosis, demyelination, or hemorrhage.

In the past few years, helical (spiral) CT scanners have been introduced into clinical practice. This machine, as compared with a conventional CT scanner, employs continuous x-ray tube rotation while the patient is moved by the motor-driven scanning table through the aperture in the scanner gantry. This volumetric data acquisition scheme is more flexible than so-called single slice scanning, because it allows multiplanar reconstructions of adjustable spatial resolution. However, contrast resolution, which is most important for delineating the thecal sac and other neurologic structures is, if anything, inferior to conventional single slice scanning. Therefore, helical spinal CT is most helpful in imaging intrinsically high-contrast bony structures and their related pathologic processes (e.g., fractures or degenerative arthritic osteophytic spurs). Helical CT scanning should not be viewed as a substitute for MRI. Nevertheless, if plain films are inconclusive, patients suspected of having acute spinal fractures are examined by helical CT, as opposed to MRI, because CT provides optimum bone detail and permits very rapid scanning of these patients. An even faster CT scanner has now been introduced, using multiple x-ray beams and detectors that create simultaneous, staggered helical scans. Such a device is capable of scanning the entire spine in about 20 seconds, albeit with the limited spatial and contrast resolution issues noted previously.

At present, nearly all patients with back pain requiring diagnostic imaging are most efficaciously studied by MRI. MRI should be obtained when patients either have failed a period of conservative treatment for back pain, or are suspected of having an illness needing urgent treatment (these would include spinal cord compression from a tumor or an infection such as osteomyelitis, diskitis, or epidural abscess). If an acute fracture is clinically suspected, plain films are first obtained. If they are not diagnostic, helical CT with sagittal and coronal reformatted images is used to provide maximum bony detail. However, intrinsic spinal cord injury, including a contusion or hemorrhage, is best depicted by MRI. From this discussion, it should be apparent that it is my opinion that plain film spinal radiography is overused as a screening tool, especially in light of its considerable insensitivity in delineating neoplasms and infections. Other than as an imaging “triage” for acute spinal trauma, plain film examination of the spine should be curtailed, using logic analogous to the abandonment of skull radiographs for evaluation of most neurologic disease. These imaging recommendations have been formulated into an algorithm.

Supplementary CT may be helpful when MRI is equivocal. This situation can occur particularly in the postsurgical spine, where metallic internal fixation devices may severely degrade or even preclude MRI, despite the use of the aforementioned HASTE sequences. With CT, usable images may still be obtainable, obviating myelography or CTM. Additionally, CT can more reliably detect calcific or ossific lesions than MRI. For example, cases of ossification of the posterior longitudinal ligament can be difficult to visualize on MRI, but are clearly depicted by CT (Fig. 8-4). MRI, however, shows the relationship of the spinal cord to the ligamentous ossification more clearly than CT. As such, I view CT and MRI as being complementary techniques.

In my experience, it is now rarely necessary to perform CTM. CTM is reserved for those patients who are unable to undergo MRI (Table 8-1) or in whom MRI and plain CT are nondiagnostic. In the unoperated patient, conventional myelography or CTM is rarely a primary diagnostic imaging study at the medical center where I practice. Only severely scoliotic patients may be requested to undergo myelography for orthopedic surgical planning, but this requirement is by no means universally applied at this institution. Myelography or CTM carries the risks of a potentially fatal anaphylactic contrast reaction, introduction of septic material into the subarachnoid space, as well as disabling post-lumbar puncture headache. It is my contention that MRI, as compared with myelography or CTM, offers superior imaging of both extra- and intradural structures with no known patient morbidity. Both MRI and CT directly visualize the extradural space, which harbors the most frequent pathologic processes: disk disease and facet joint arthropathy. Myelography, which fills the intradural compartment (thecal sac) with contrast, at best provides indirect evidence of extradural pathology by virtue of indentations on the sac or nerve root sleeves. Lesions in the extradural space that do not adjoin the thecal sac or the nerve root sleeves will be undetectable by myelography. No such limitation applies to MRI or CT. MRI pulse sequences have been developed that permit the rapid acquisition of heavily T2-weighted images with high spatial resolution. These are called fast spin echo or turbo spin –echo. With strong T2 weighting, the cerebrospinal fluid appears bright, and essentially identical in appearance to intrathecal contrast used in CTM (Fig. 8-5). Thus, the CTM image equivalent is achieved noninvasively by MRI, which, in my opinion, usually renders CTM superfluous.

As noted previously, T2 signal abnormalities in the spinal cord can represent several pathologic processes, including gliosis, edema, or tumor. The differentiation between these entities has been refined through the use of the MRI contrast agent, gadolinium diethylenetriamine pentaacetic acid (DTPA). Gadolinium, a rare earth metal, is a paramagnetic element, causing shortening of the T1 relaxation time of tissues that accumulate it. T1 shortening translates into brightening on a magnetic resonance image that is T1 weighted. Moreover, in humans, the transport kinetics of gadolinium are identical to those of iodinated radiographic contrast agents.

This means that both gadolinium and iodinated contrast material cause enhancement primarily through accumulation of either substance in the extravascular soft tissues. For delivery of gadolinium, the metal is chelated to a carrier molecule, DTPA, and this pharmaceutical is injected intravenously. Formerly, dosage recommendations were 0.1 mmol/ kg body weight. Presently, most patients are satisfactorily imaged with a 10-mL dose. Gadolinium carries almost no patient risks, with only extremely rare instances of anaphylaxis, far fewer than with iodinated contrast used for CT. Even patients with hepatic and renal compromise can tolerate gadolinium DTPA more readily than iodinated contrast. Nevertheless, it is prudent to administer gadolinium with discretion, particularly in these debilitated patients. Gadolinium will cause enhancement of most spinal tumors, as well as postoperative granulation tissue and many inflammatory processes.

Low back pain is often a nonspecific symptom that may relate to paraspinal pathology involving retroperitoneal structures, including tumors. Therefore, when imaging the spine with either CT or MRI, it is recommended that at least some coverage of the paraspinal region be provided, particularly in the axial sections. Pathology such as aortic aneurysms, infections, and retroperitoneal tumors can be demonstrated via either technique.1 Secondly, it is extremely important that the radiologist be provided with pertinent clinical history, particularly with regard to any localization, especially lateralization of pain, weakness, or other neurologic symptomatology. The imaging facility with which I am affiliated has patients note their perception of pain on an anatomic diagram (Fig. 8-6), as well as answer a series of questions concerning their symptoms. This is a procedure that I have found very useful on a number of occasions, as it is well known that many imaging abnormalities may have no correlative clinical significance.

In recent years, there has been a proliferation of so-called open magnet imaging devices. These machines have great marketing appeal, and they are often billed as being “patient friendly” and capable of accommodating even those who are claustrophobic or morbidly obese. These claims are not unwarranted. However, it should be recognized that even recently designed open systems cannot deliver image quality commensurate with that of high-field systems. This is readily explained when it is understood that the open system magnet field strengths range from one fifth to one tenth those of a high-field imager. The lower field strength means that the resultant images will be derived from proportionately weaker radio signals coming from the patient. Although novel pulse sequences such as constructive interference in the steady state (CISS) can help to alleviate this problem, as is usually the case in MRI, this comes at a price: reduced T2 sensitivity (Fig. 8-7). However, low-field scanners can produce excellent cervical spine scans, as this region of the body is relatively diminutive. The small diameter of most necks permits the receiving coil to be close to the region of interest (cervical spine), maximizing the ability to detect signals arising from the spinal tissues. Additionally, low-field units are less susceptible to artifacts arising from vascular pulsation or respiration, both of which can degrade neck images. Nevertheless, in general, low-field scanners require more time to scan the patients, again a function of their operating with weaker magnets. Longer scan times may not be tolerated by a patient as a result of pain, with resultant artifacts secondary to patient motion. Low-field imagers really cannot compete with their high-field brethren in the production of thin section scans, where signals are at a premium. Therefore, it is my opinion that “all imagers are not created equal.” Whenever possible, I prefer to scan patients with high-field machines to guarantee optimal image quality. Newer high-field units have a so-called short (magnet) bore, with flared ends of the scanning aperture, soft lighting, and abundant ventilation. Such amenities significantly diminish the patient’s perception of confinement. Where necessary, the nursing staff can provide either oral or intravenous conscious sedation, which usually makes it possible to scan even the most anxious individual.

The most common causes of neck and back pain that can be imaged are intervertebral disk disease and facet joint degeneration. Disk disorders are more prevalent in adulthood, whereas facet degeneration is more frequent in the elderly. However, the two disease processes are often concurrent. Although MRI depicts disk pathology with great sensitivity, recent publications have noted the detection of disk abnormalities in asymptomatic individuals.2,3 Therefore, any radiologic finding of disk disease must be viewed in the context of clinical symptomatology before it should be deemed significant. Furthermore, some types of disk pathology occur more frequently in the asymptomatic patient. For example, Jensen and colleagues3 observed that a bulging disk, defined as a concentric, uniform expansion of disk material (Fig. 8-8), was more frequently asymptomatic than what was termed an extruded disk. (These authors grouped herniated disks in the category of extruded disk. This apparent terminologic variation underscores the need for a more standardized nomenclature of disk abnormalities, to predictably assign clinical significance.) A protruding disk is a focal deformity of the outer disk margin, with a broad base relative to the anteroposterior (AP) extent of the lesion (Fig. 8-9). This may also be associated a tear of the outer disk material, the anulus, and thus is referred to as an anular tear (Fig. 8-10). The pathology of an anular tear is analogous to a tear of a meniscal cartilage. It usually presents as an area of increased T2 signal in the anulus, highlighted from the dense collagen of the normal anulus, which is dark on T2-weighted scans. A herniated disk is a focal distortion of disk margin, in which the AP extent of the deformity approaches or exceeds its width (Fig. 8-11). Extruded disks, as defined here, remain connected to the “parent” disk while progressing cephalad, caudad, or bidirectionally relative to the plane of the disk space (Fig. 8-12). Finally, a free disk fragment is separable from the parent disk (Fig. 8-13).

In the cervical spine, midline posterior disk protrusions or herniations are clearly defined by sagittal MRI scan series (Fig. 8-14) as the section plane is perpendicular to the site of the pathology. T2-weighted images can show increased signal intensity within the compressed spinal cord, representing either edema or gliosis.4 (High T2 signal could also occur in a concurrent abnormality, such as a spinal cord tumor. Gadolinium-enhanced MRI scanning has improved the detectability of spinal cord neoplasms. Typically, these lesions enhance whereas the edematous component does not.) A laterally situated cervical disk herniation, typically causing radiculopathic symptoms, is optimally visualized by MRI in the axial plane (Fig. 8-15). The parasagittal images are not as revealing in this location because of the oblique orientation of the cervical neural foramina. However, a lateral, intraforaminal lumbar disk herniation can also be clearly seen with either axial or sagittal MRI, owing to the parasagittal orientation of the plane of the lumbar neural foramina. As previously noted, a portion of disk can separate from the parent disk, producing a free fragment.5 Virtually all free fragments remain extradural in location. Rare cases of fragments perforating the posterior longitudinal ligament and dura and, thus, lodging intradurally have been reported,6 but almost always diagnosed retrospectively at surgery (Fig. 8-16). The conus medullaris is well delineated on routine sagittal scans, permitting exclusion of a tumor at that site. On occasion, a conus medullaris tumor can symptomatically mimic a disk herniation.

Although MRI provides superlative depiction of disk disease, axial CT is often just as effective, particularly when there are associated osteophytic spurs. As previously noted, CT provides excellent delineation of bone. Clinical localization is helpful in directing the CT examination to the appropriate levels of the spine because it is impractical to scan the entire spine axially by CT. “Surveys” of this sort are best handled by MRI. However, the more directed an examination of any type can be, the more likely it is to provide useful clinical information.

Thoracic disk pathology was formerly thought to be much less prevalent than those involving lumbar disks. However, the incidence was underestimated because of low imaging sensitivity of plain films and myelography, particularly in the thoracic spine owing to the typical hyphotic curvature, which limited pooling of myelographic contrast agent along the posterior disk margins. The ease of comprehensive thoracic spine imaging by MRI and its abundant use has altered our perception of thoracic disk disease (Fig. 8-17). Thoracic disks have a higher frequency of dystrophic calcification than disks in other regions of the spine. As a result, CT may be able to delineate the disk abnormality, as this modality is quite sensitive to calcification.

In recent years, serial spinal MRI has validated the merit of conservative treatment strategies. Disk herniations, particularly those that are large, can occasionally regress over time without surgical intervention (Fig. 8-18). Yock7 has offered the explanation that the herniation may have associated hemorrhagic elements, which could be progressively resorbed and thus account for the apparent reduction in size. Spontaneous epidural fibrosis, which would contain the herniation, likely plays a role in the healing process, promoting reduction in size of the herniation.

In the past, some of the most difficult issues to be resolved have related to postoperative cases. The determination of whether there is a recurrent or residual disk herniation, or both, has been simplified with gadolinium-enhanced MRI.8 Gadolinium will enhance postoperative scar tissue, because of the presence of vasa vasorum within the fibrosis. Disk material, not possessing this vascular component, will not undergo enhancement, provided scanning is accomplished promptly after gadolinium is injected. Otherwise, the disk material may “imbibe” the contrast agent. Obviously, differentiation of scar from residual or recurrent disk disease is critical in the surgeon’s decision as to whether reoperation would be beneficial. I have also found that supplemental axial T2-weighted imaging can be of help in distinguishing a disk fragment from an entrapped nerve root sleeve (Fig. 8-19). Ross and colleagues9 noted that over a period of a year, postoperative epidural scar underwent little change in morphology. In another paper, Ross and associates10 found that there was a correlation between the intensity of postoperative radiculopathic symptoms and the quantity of scar. Lastly, Ross and coauthors11 described a pattern of gadolinium enhancement of the endplates typical of benign fibrosis in the post-diskectomy patient. This pattern of enhancement was not indicative of postoperative diskitis, an important differential consideration when reviewing MRI scans of this patient population.

Over the past few decades, diskography has been advocated, principally by the orthopedic surgical community as a methodology for detecting and characterizing disk pathology via injection of contrast material into the disk and imaging with plain radiography or CT. It has also been used as a provocative test (diskomanometry) to observe whether injecting the disk with saline reproduces the patient’s symptoms, implying causation. Publications endorsing the utility of diskography have appeared, again mostly in the orthopedic literature.12–14 However, contrary opinions concerning the efficacy of diskography have been cited.15 There has been limited endorsement of diskography by radiologists, particularly in view of the large accumulated experience with MRI and CT, which are noninvasive techniques. Diskography carries the risk of introduction of infection into the disk and adjacent bone (Fig. 8-20), as well as potentially life-threatening allergic reaction to the contrast agent.

Facet joint degeneration occurs most commonly in the lumbar and cervical portion of the spinal column and least commonly in the thoracic region. Presumably, this is because of the limited mobility within the thoracic spine, compared with cervical and lumbar sites. Degenerative arthritis leads to facet joint space narrowing and eburnation. Bony spurs may produce thecal sac or nerve root sleeve compression, secondary to central spinal canal or neural foraminal stenosis. This spinal stenosis can be aggravated by thickening of the ligamentum flavum posteriorly, an abnormality that frequently accompanies facet joint degeneration. Moreover, facet arthropathy often accompanies degenerative disk disease. The two pathologic processes together may cause spinal stenosis, the extent of which can be accurately demonstrated by MRI (Fig. 8-21). CT is an excellent alternative imaging study, provided the clinical examination can serve to accurately direct the placement of the axial scans to the suspected interspace level.

As facet joint degeneration intensifies, spondylolisthesis can occur, with the resultant malaligned vertebrae aggravating the spinal stenosis. In nearly all cases, the vertebral segment that displaces does so anteriorly relative to the contiguous, caudal segment. The spondylolisthesis can be visualized by plain radiographs, MRI, or CT (Fig. 8-22). Degenerative spondylolisthesis should be distinguished from spondylolisthesis with an accompanying pars defect, known as spondylolysis. The defect is most readily depicted by sagittally reformatted CT, and is most commonly seen at the L5 level (Fig. 8-23). Spondylolisthesis can occur secondary to spondylolysis. In this case, the greatest stenosis is seen at the neural foramina, secondary to the pars above the defect being displaced into the foramen. The central spinal canal, if anything, is elongated in its AP dimension owing to the effective break in the neural arch secondary to the spondylolysis. However, some central canal stenosis can ensue as a result of encroachment by eburnated bone arising from the pseudoarthrosis created by the spondylolysis. Nearly always, the spondylolysis is bilateral.

Neoplastic disease in the spine usually presents with back pain. The most common tumor is a metastasis from a breast or lung carcinoma. MRI is exquisitely sensitive to the presence of bony vertebral metastatic disease16–21 and any associated epidural tumor extension that is causing spinal cord or nerve root compression (Fig. 8-24). In my institution, myelography is no longer performed on patients to exclude cord compression from metastatic disease, because MRI provides superior delineation of this pathologic process noninvasively, and with a degree of precision more than sufficient to plan radiotherapy. Metastatic disease involving the leptomeninges can also be detected on MRI, particularly with gadolinium enhancement (Fig. 8-25). However, even with enhancement, MRI is only about one third as sensitive as serial lumbar punctures in detecting leptomeningeal spread of tumor.22

Patients with suspected metastatic disease who are unable to be examined with MRI for technical or medical reasons can be evaluated by CT, after directing the study to a specific region of the spine by the aid of the clinical findings, plain radiographs, and, if time permits, a radionuclide bone scan. This latter study is still a very sensitive and inexpensive method of surveying the vertebral column for the presence of bony malignancy, although purely lytic metastases may not be detectable. As helical scanning has become more readily available, more extended CT scans of the spinal column are now possible, although with limited spatial and contrast resolution as compared with MRI.

Intradural neoplasms can be classified as intramedullary or extramedullary. Intramedullary tumors are most commonly low-grade astrocytomas or ependymomas. Spinal cord gliomas can involve virtually the entire extent of the spinal cord. Again, MRI is the imaging modality of choice. These neoplasms usually expand the spinal cord and alter its signal intensity, making their visualization easy. However, it is often possible to differentiate the tumor margins from accompanying cysts or edema by using gadolinium-DTPA23 (Fig. 8-26). Most extramedullary intradural tumors are meningiomas or schwannomas. These neoplasms may compress both the spinal cord and nerve roots (Fig. 8-27), findings easily delineated by MRI. However, distinction between meningioma and schwannoma can sometimes be accomplished with CT, which can better demonstrate calcification seen occasionally but almost exclusively with meningioma. Spinal schwannomas can occur as solitary lesions or as a manifestation of neurofibromatosis I, an inherited neurocutaneous disorder, or phakomatosis.24

Some intramedullary tumors may be associated with syringomyelia.25 If the detection of one pathologic process in the spine has been facilitated by MRI, it is syringomyelia (Fig. 8-28). Formerly, such time-consuming, invasive procedures as gas myelography and, later, CTM were the only methods for demonstrating cysts within the spinal cord. CTM was not universally successful in demonstrating a syrinx, as this necessitated progressive imbibition of contrast agent into the central cystic cavity. Chiari malformations of the hindbrain and cervical spine with cerebellar tonsillar displacement through the foramen magnum and an accompanying syrinx are optimally demonstrated by MRI.26 The syrinx can extend the length of the entire spinal cord. In the lumbar spine, dysraphic states, such as meningoceles or tethering of the spinal cord with an associated lipoma, are best imaged by MRI27 (Fig. 8-29).

Demyelination in the spinal cord can present as a radiculopathy28 clinically mimicking disk herniation, particularly in the cervical region. Before MRI, there was no method capable of directly identifying demyelinating plaques within the spinal cord, although frequently such lesions could be demonstrated in the brain by CT. Typically, demyelination produces one or more discrete foci of elevated T2 signal intensity within the spinal cord, at times with cord expansion if the lesion is acute (Fig. 8-30). Besides the clinical features, confirmation that the signal abnormalities in the cord may be caused by a demyelinating process can be obtained by MRI of the brain, demonstrating additional foci of T2 hyperintensity in the periventricular white matter.29

Trauma to the spine traditionally has been initially imaged with plain x-ray films. These studies often serve to guide subsequent CT scanning, which provides excellent imaging of the extent of most fractures. MRI provides superior definition of soft tissue injuries, especially if there is an associated spinal hematoma or cord contusion.4,30 However, in the acutely traumatized patient who may be in traction or on external life-support systems, it is often physically impractical to employ MRI. The recent introduction of wide-bore, so-called open magnet systems may facilitate scanning of these patients by granting support personnel greater access to the patient. Nevertheless, as previously noted, image quality issues may outweigh ease of access (Fig. 8-31).

An increasingly common cause of back pain is spinal osteomyelitis and epidural abscess, which are now being seen with greater frequency because of larger numbers of intravenous drug abusers and generally debilitated patients. Vertebral and intervertebral disk destruction is the critical observation in diskitis or osteomyelitis, yet such abnormalities occur late in the course of the disease. With epidural abscess, plain films will be unrevealing. Therefore, plain films should not be relied on to rule out an infectious process. MRI is the preferred technique for imaging both osteomyelitis or diskitis and epidural abscess31,32 (Fig. 8-32). Because epidural abscesses may extend in a discontinuous fashion, it is recommended that the entire spine be imaged, at least in sagittal plane, with supplemental multiplanar sequences in the area of clinical symptomatology. An epidural abscess may take the form of a pus-filled cavity. In this case, the membrane delimiting the cavity will enhance, whereas the liquefied center will not. Another form of abscess, known as a phlegmon, tends to enhance more uniformly.

Perhaps the rarest cause of back pain that can mimic symptoms of spinal stenosis is a vascular malformation involving the spinal cord or surrounding dura mater.33–37 In preparation for either surgical resection or therapeutic embolization of a vascular malformation, selective spinal angiography is an essential diagnostic and therapeutic procedure. However, MRI has supplanted myelography in detection of intramedullary vascular malformations and, often, dural-based arteriovenous fistulas (Fig. 8-33). Formerly thought to be a rare phenomenon, the dural arteriovenous fistula can mimic the symptoms of spinal stenosis. It is imperative to search for such a lesion, particularly in the middle-aged patient, as failure to detect and treat this lesion can lead to progressive para- or quadriplegia secondary to venous infarction of the spinal cord.

In most patients suffering from back pain, MRI is the preferred spinal imaging procedure. Nevertheless, this powerful, but expensive tool should be used with discretion, with both reasonable clinical indications for obtaining the study and, where possible, only after failure of conservative treatment strategies. Remembering that most back pain resolves over time will save many patients an unnecessary imaging study and society at large a considerable savings of valuable health care resources.