The most frequently encountered finding was spinal fluid collections, which we saw in 85% of the patients in this series, compared with the 67–100% reported in other series [10–12, 19]. The collections tended to be nonfocal and often extended over five or more spinal segments. Axial T2-weighted images were particularly useful for visualizing the dura deep to the collection, signifying an epidural location (Fig. 1A, 1B).

The mechanism of development of the spinal collections is not known. One explanation may be direct leakage of CSF into the epidural space with tracking of fluid through the space. Another theory is transudation of intravascular fluid from hyperemic meninges or engorged veins across a hydrostatic pressure gradient [15]. In the surgical cases in our study, the patients had extensive collections that extended to regions remote from the surgically documented site of leakage (Fig. 2). Although some of these collections may seem to be explained by the transudation theory, at least one patient had direct extension to C1–C2 through a dural defect or fistula.

Dilated epidural veins (Fig. 3A, 3B) were found in 77% of the patients in this series, compared with 67–88% in other series [10, 11, 15]. Enlargement of epidural veins has been described previously in association with both spontaneous intracranial hypotension [20–23] and hypotension after lumbar puncture [24]. Enlarged veins also have been reported as potential mimickers of a dural arteriovenous fistula [23]. The presence of dilated veins may be explained by the Monro-Kellie doctrine, which states that as CSF volume decreases, blood volume increases in an attempt to maintain a stable pressure, essentially leading to compensatory vasodilatation in the brain, spine, and meninges [1, 12, 15].

In the assessment of the epidural venous plexus, both morphologic and size characteristics were used. The size criteria used were subjective, which is a potential weakness of this study. However, to our knowledge, no accurate measures of plexus thickening have been established in the literature. The morphologic criterion used was the festooned or light-bulb appearance of the plexus, which has been described by Chiapparini et al. [10]. The festooning results when the enlarged plexus causes collapse of the lateral aspects of the thecal sac while the sac maintains its midline attachment to the posterior longitudinal ligament (Fig. 4A, 4B).

According to the Monro-Kellie principle, dural enhancement is assumed to result from dural vasodilation and engorgement. Dural enhancement was assessed in the 11 patients who underwent contrast-enhanced spinal examinations and was seen in 73% of the patients, compared with 66–86% in other series [10, 12]. The enhancement tended to be smooth and circumferential (Fig. 5) and was often but not always accompanied by intracranial dural enhancement. Three patients in the study lacked intracranial dural enhancement but did have spinal dural enhancement. Mokri et al. [9] reported lack of intracranial pachymeningeal enhancement in several patients with symptomatic spontaneous intracranial hypotension and suggested that the quantitative volume losses and pressure changes might not have been sufficient to induce the intracranial dural engorgement and enhancement.

There have been several reports [25–28] of an association between spontaneous intracranial hypotension and structural abnormalities. Association of spontaneous intracranial hypotension with nerve root cysts, pseudomeningocele, disk herniation, and transdural osteophytes has been reported. The abnormalities seen in our study included dural defects, dorsal arachnoid diverticula (Fig. 6A, 6B), and nerve root cysts. A minority (31%) of patients in this series had imaging findings of identifiable structural abnormalities in association with the site of leakage, compared with 6–35% in other series, in which conventional noninvasive spinal MRI was used [10, 11, 19].

In two studies [18, 29] in which mainly invasive imaging such as MRI, CT myelography, or nuclear cisternography was performed, higher rates of association (58–69%) between structural abnormalities (meningeal diverticula) and spontaneous intracranial hypotension were found. In those studies, conservative therapy such as bed rest, caffeine intake, and blood patching had already failed, a factor that might have skewed the results.

The data from the two studies [18, 29] suggest that conservative therapy is more likely to fail in the care of patients with spontaneous intracranial hypotension with structural defects than it is in the care of patients without these defects. Likewise, we found that two of our three patients with imaging findings of structural abnormalities underwent unsuccessful conservative treatment, and they needed reparative surgery for cure. Analysis of a larger number of patients certainly is required for statistical proof of this finding. Hyun et al. [17] also found a significantly increased likelihood of failure of conservative treatment of patients with direct signs of leakage at nuclear cisternography. It is important to note that if a patient does not have symptoms, the presence of a structural abnormality alone has not been proved to indicate any increased risk of development of a leak.

One finding that deserves special mention is the C1–C2 sign. This sign, which has also been called the C1–C2 false localizing sign [30, 31], is seen on MR images as a focal area of fluidlike signal intensity and on CT myelograms as a CSF collection between the spinous processes of C1 and C2 (Fig. 7A, 7B). More than one half of the patients in this series had this sign, suggesting a clear association. In other series, this sign has been found 33–67% of the time [15, 19]. It should be noted that this fluid collection does not necessarily denote the site of CSF leakage. Two of our four surgical patients also had the C1–C2 sign. The first was found to also have leakage in the thoracic spine at myelography that was confirmed at surgery. The other patient had contrast leakage at C1–C2 extending into the area of the right brachial plexus at myelography, but at surgery the leakage was not found at C1–C2 and instead appeared to originate from below that level.

These two cases exemplify how the C1–C2 sign can incorrectly localize the site of leakage, sometimes even when contrast leakage is perceived at C1–C2 at myelography. This sign also has been reported in association with intracranial hypotension after lumbar puncture [15], which further supports the assertion that retrospinal fluid at C1–C2 does not represent direct CSF leakage. If the fluid at C1–C2 is not from direct leakage, it may be from indirect leakage. Schievink et al. [30] postulated that these collections are true CSF collections that occur when fluid from a spinal epidural collection travels rostrally and escapes from the epidural space into the retrospinal soft tissues. The C1–C2 level may be prone to the escape of fluid because of lack of osseous support, mobility of the segment, lack of epidural fat, and laxity of the connective tissues at this level. In our series, six of the seven patients with the C1–C2 sign also had epidural collections.

A number of patients in this series lacked some of the typical intracranial findings. One patient had completely normal cranial and spinal MRI findings but had abnormal CT myelographic findings. Five patients lacked cerebellar tonsillar depression, and these patients did have multiple spinal findings, including the C1–C2 sign and fluid collections. Three patients lacked intracranial dural enhancement but had dural enhancement in the spine, fluid collections, dilated epidural venous plexus, and dilated veins. This finding suggests that spinal imaging may have a role in cases in which intracranial findings are absent or equivocal. Results of other studies also have shown the absence of intracranial findings in 10–28% of cases [7, 11, 29].

Axial T2-weighted fat-suppressed images were useful for detection of fluid collections, vein enlargement, and venous plexus enlargement. We found sagittal T2-weighted fat-suppressed images helpful for evaluation of the overall extent of epidural fluid collections and epidural vein enlargement. The C1–C2 sign is often subtle and is best visualized on sagittal fat-suppressed T2-weighted images. If some form of fat-suppressed T2-weighted sagittal image is not obtained, this sign is likely to be underrecognized. Axial and sagittal contrast-enhanced T1-weighted images were of great value for visualizing dural enhancement venous plexus enlargement. An imaging algorithm based on our experience and findings in the literature is shown in Figure 8.

We identified an array of spinal findings of spontaneous intracranial hypotension. These findings were identified in all 13 patients in this series, compared with 88–94% in other series [10, 11, 19]. Specifically, fluid collections, thickened epidural venous plexus, dilated veins, and dural enhancement were present in most of our patients and have been consistently encountered in other series. Though analysis of a larger number of cases is required to determine whether these findings are truly characteristic of the syndrome, it appears that spinal imaging will likely show one or more findings in patients with spontaneous intracranial hypotension. We therefore advocate imaging of the spine with contrast-enhanced sagittal T1-weighted, sagittal and axial T2-weighted fat-suppressed, and unenhanced sagittal and axial T1-weighted fat-suppressed sequences in the care of patients with suspected spontaneous intracranial hypotension, particularly when the intracranial findings are absent or equivocal and when surgery is planned. The C1–C2 false localizing sign is another important finding. Visualization of this finding can lead to the radiologic diagnosis of spontaneous intracranial hypotension. It is also important to realize that this finding is not necessarily indicative of the leakage site.