Routine brain MRI protocols consisting of, at a minimum, contrast-enhanced 3D T1-weighted imaging, T2-weighted imaging, FLAIR, and susceptibility weighted imaging, are sufficient to evaluate for intracranial sequelae of SIH. Such sequelae include diffuse dural thickening and enhancement, subdural collections, pituitary engorgement, dural venous sinus engorgement, sagging of the brainstem and cerebellum, and superficial siderosis [1] (Fig. 2). The presence of any one of these imaging findings in a patient with an orthostatic headache constitutes sufficient evidence to warrant further investigation for underlying spinal CSF leak or venous fistula.

As most CSF leaks causing SIH originate in the spine, the work-up of patients with suspected SIH must include whole-spine imaging to evaluate for the presence of an extradural CSF collection [4]. Although routine spine imaging is usually sufficient to identify a dorsal or ventral epidural fluid collection, CSF leaking laterally via a perineural cyst or lateral dural defect may not be readily detected on conventional spine MRI protocols comprising 2D sequences without fat saturation. Therefore, inclusion of a 3D fat-saturated (FS) heavily T2-weighted imaging sequence is warranted to evaluate not only the epidural space but also the paraspinal soft tissues for the presence of extradural CSF [5]. Accordingly, our protocol includes a coronal 3D fat-saturated HASTE sequence with an echo train length (ETL) of 256, isotropic resolution of 0.9 mm³, and TR/TE of 8000/271. The sequence is reformatted in sagittal and axial planes as well as into a coronal volumetric maximum-intensity-projection image. In addition, a sagittal STIR sequence is obtained at each spinal level. The entire protocol encompassing the cervical, thoracic, and lumbar spine is acquired in approximately 40 minutes.

The 3D FS T2-weighted imaging sequence is useful not only for identifying extradural CSF but also for characterizing meningeal diverticula, which can be a source of CSF leak and a nidus for CVF [13]. The presence, complexity, and laterality of meningeal diverticula can inform patient positioning for later decubitus dynamic myelography, and careful scrutinization and comparison between the contours of meningeal diverticula on 3D FS T2-weighted imaging and those on CT or MR myelography often lead to detection of a leaking meningeal diverticulum (Fig. 3).

Dynamic myelography can be performed via a fluoroscopic or CT technique, with each modality having strengths and weaknesses. Fluoroscopic dynamic myelography is usually performed with digital subtraction (i.e., digital subtraction myelography [DSM]), and a tilt table and biplane may be used to improve leak detection. However, because of respiratory motion artifacts on digital subtraction images in the thoracic spine, where most CVFs occur, general endotracheal anesthesia is often used to control breathing and minimize artifacts [14]. In addition, although DSM shows excellent temporal resolution, its restricted FOV limits evaluation to the areas specifically examined by the operator, thus necessitating pretest suspicion of leak location.

We have had good success in the use of dynamic CT myelography (CTM) for localization of CVF and dural defects. Dynamic CTM offers unique advantages to DSM, including excellent spatial localization, acquisition of whole-spine images, and simultaneous characterization of surrounding anatomy for treatment planning purposes. In addition, the patient does not need to be placed under general endotracheal anesthesia to avoid breathing artifacts. However, these benefits are at the expense of a higher radiation dose and poorer temporal resolution compared with DSM. Both DSM and dynamic CTM remain the tests of choice for the detection of CVF, and future studies should seek to quantify and compare the relative sensitivities of the two modalities.

Patient positioning during dynamic CTM depends on the type of leak suspected. If initial spine MRI shows a ventral epidural collection, then the patient should be placed in the prone position to localize the site of the ventral dural defect. If spine MRI does not show an epidural collection, then the patient should be placed in the lateral decubitus position to study the lateral dura and meningeal diverticula. Performing the procedure in two separate sessions is often necessary to carefully examine both sides during early dynamic contrast-enhanced phases.

Unlike DSM, in which a tilt table may be used to facilitate cranial passage of contrast material, dynamic CTM requires the use of an extrinsic device for this purpose. Some institutions place a foam wedge under the patient's hips [15, 16]. Although this maneuver allows straightforward dynamic imaging, it precludes accurate measurement of opening pressure. We have achieved good results using an inflatable mattress (HoverMatt Air Transfer System, HoverTech International) for dynamic positioning.

Regardless of the modality used, all dural punctures in patients with suspected leaks should be performed with a noncutting sidehole needle, which results in significantly fewer postpuncture headaches compared with traditional cutting spinal needles [17].

Before the procedure is performed, spine MRI should be reviewed to assess the location of the epidural fluid collection. Osteophytes, which are potential sources of dural tears, should be noted. The suspected level of the defect (and thus the distance of the defect from the lumbar puncture) informs how long the patient's hips should be raised to allow cranial passage of intrathecal contrast material. For example, a suspected leak in the lumbar spine requires less time with the hips raised (achieved by the inflation) than does a suspected leak in the cervical spine.

The inflatable mattress should be folded in half and positioned on the CT table in proximity to the CT gantry. The patient should be positioned prone on the table with the mattress under their hips. A test inflation should be performed to ensure that the hips are adequately raised above the head to facilitate the rapid cranial passage of contrast material.

During planning for initial lumbar puncture, a whole-spine scout image should be obtained. This scout image allows rapid whole-spine imaging to be performed later in the procedure without repeating the scout acquisition. During lumbar puncture, the lowest spinal level possible should be accessed to allow the greatest number of spinal levels to be examined cranial to the level of access. During lumbar puncture, care should be taken to attempt to position the tip of the needle as close as possible to the center of the thecal sac, to avoid a subdural injection. Once CSF returns through the needle, approximately 0.2 mL of contrast material should be injected to confirm subarachnoid positioning. Subsequently, 5–10 mL of iodinated contrast material (iopamidol, 300 mg I/mL; Isovue-M 300, Bracco) should be instilled through connective tubing, followed by re-placement of the stylet and needle removal. The mattress should be inflated at this time. Depending on the suspected level of leak and the degree of lumbar lordosis or spinal stenosis, the time of inflation may range from 10 to 20 seconds. Subsequently, the mattress should be deflated and whole-spine scanning rapidly performed.

As images appear on the scanner console, they should be scrutinized in real time for contrast material layering along the ventral thecal sac across the entire spine. Ideally, an abrupt transition point will be identified where contrast material moves from the subarachnoid space to the epidural space (Fig. 4). If contrast material has not extended sufficiently in the cranial direction after the first scan is obtained, scanning should be aborted, the mattress reinflated for an additional 10–20 seconds, and the scan repeated.

CVFs present a particular diagnostic challenge because of the rapid intermittent nature of their leakage. Thus, in addition to rapid early phase imaging, authors have suggested the use of provocative maneuvers to aid CVF detection. For example, Caton et al. [7] showed the safe, well-tolerated use of pressure augmentation with sterile saline to both enhance detection of CVF and allow calculation of compliance curves to characterize underlying leak pathology. In addition, Amrhein et al. [8] showed that imaging during the inspiratory phase increases the conspicuity of CVF. Therefore, we use both of these maneuvers during dynamic CTM to increase sensitivity for CVF.

Before dynamic CTM is performed, spine MRI should be reviewed to identify the presence and laterality of perineural cysts, which can serve as both a nidus for CVF and a site of primary leakage. The presence of large or irregular perineural cysts on one side of the spine or an asymmetric burden of perineural cysts might guide the choice of the decubitus side on which the patient should initially be positioned. If the first dynamic myelogram is normal, then the patient returns on a separate day to have a repeat dynamic myelogram obtained with the contralateral side down to complete the workup.

The patient is positioned on the table with the mattress folded in half under their hips. A test inflation of the mattress is performed to ensure that positioning facilitates the cranial passage of contrast material as well as to ensure that the patient is safely centered on the mat and will not roll forward or backward during the procedure. During initial planning, a whole-spine scout image should be obtained. Access to the thecal sac should be obtained in a manner similar to that previously described. Once CSF returns through the needle, opening pressure should be measured with a digital manometer (Compass Digital Manometer, Centurion Medical Products). As long as the opening pressure is either normal or low, infusion of sterile saline may be initiated via incremental 5-mL aliquots, measuring and recording the pressure after each aliquot. Once the pressure reaches approximately 25–30 cm of H₂O, the infusion is stopped. A total of 5–10 mL of iodinated contrast material (300 mg I/mL) is then instilled, the stylet re-placed, and the needle removed. The mattress is then inflated for approximately 10–15 seconds, after which time it is deflated and rapid scanning is initiated. As they appear on the scanner console, the images should be scrutinized for contrast material layering dependently in the lateral thecal sac, filling any perineural cysts, and extending at least to the craniocervical junction. Two whole-spine acquisitions should be obtained in immediate succession. The patient may subsequently be repositioned to the opposite lateral decubitus position, and one final acquisition performed (Fig. 5).

Intrathecal use of gadolinium-based contrast agent for MR myelography, although off label, is considered a safe and effective method for detection of CSF leaks [10]. The literature suggests that slow leaks, which may not be evident on early phase imaging, may be optimally detected with MR myelography [10, 18]. This difference may reflect a combination of improved tissue contrast resolution for MRI compared with CT as well as the substantial delay in imaging for MR myelography. When pressure-augmented decubitus dynamic myelography with simultaneous intrathecal injection of gadolinium-based contrast agent is performed, the pressure augmentation is anticipated to increase the conspicuity of extrathecal contrast material originating through slow or subtle leaks [11]. Using these combinations of techniques, we have detected several leaks that were occult on CTM or 3D FS T2-weighted imaging (Fig. 6).

According to our diagnostic algorithm, patients with findings of SIH on brain MRI but no extradural collection on spine imaging undergo decubitus dynamic CTM to evaluate for CVF. If decubitus CTM is normal, then the patient returns on a separate day to undergo dynamic CTM in the opposite decubitus position. During the second myelogram, 0.2 mL of intrathecal gadolinium-based contrast agent (gadobenate dimeglumine, 0.5 M [MultiHance, Bracco]) is simultaneously instilled for subsequent MR myelography, which is performed within 30–60 minutes after dynamic CTM. MR myelography is performed using a multiplanar T1-weighted sequence with a modified 3-point Dixon technique for fat saturation and with a section thickness of 3 mm, an inter-slice gap of 0.5 mm, TR of 475 ms, ETL of 3–4, TE of 10 ms, bandwidth of 50 Hz, and matrix size of 320 in the frequency-encoding direction and 224 in the phase-encoding direction. After sagittal whole-spine acquisition, real-time monitoring is performed to prescribe additional axial and coronal planes of interest, with the goal of minimizing scan time while optimizing image quality.

Epidural patching with autologous blood and/or fibrin glue— Large-volume autologous EBP is the mainstay treatment for postdural puncture headache and can be performed with or without image guidance. In addition, fibrin glue (a biologic adhesive material consisting of reconstituted human fibrinogen and thrombin) has potentially greater efficacy in symptom relief, particularly when the site of the leak is identified [21, 22]. In patients who have undergone a procedure with possible dural violation (e.g., lumbar puncture or spinal epidural anesthesia), first-line therapy should include prompt EBP at or near the level of instrumentation. The rates of symptomatic relief in these patients is high, with a clinical trial suggesting that approximately 90% of patients have improvement in symptoms after EBP [23]. In patients with suspected SIH, data suggest that efficacy may be higher for targeted, rather than empirical, treatment [21]. In addition, our experience and the published literature suggest that CVF may not respond to EBP in a sustained fashion [22]. Therefore, patients with an orthostatic headache ideally should undergo an algorithmic diagnostic workup to identify the site of leak. However, if a patient has severe symptoms, diagnostic workup should not delay empirical EBP, which can provide substantial relief while having only modest risks.

Once the site of a leak is identified or suspected, attempts should be made to direct autologous blood and/or fibrin to the site of leak. Such attempts can be particularly challenging in the context of ventral dural defects, as EBP is most commonly performed from a dorsal approach. We have successfully performed far-lateral approaches with longer spinal needles to target the ventral epidural space (Fig. 8). CT guidance is critical in these cases to delineate intervening anatomy and plan a safe needle trajectory.

Surgical repair—Patients with dural defects refractory to percutaneous therapies should undergo surgical consultation for primary dural repair. The ventral location of these defects can present a particular surgical challenge and may require an anterior approach, which may have significant associated morbidity. However, a successful posterior approach for ventral dural repair has been described and performed with great efficacy [24] (Fig. 4D).

Percutaneous fibrin glue embolization—CVF can be successfully treated using percutaneous injection of fibrin glue with CT guidance, by targeting the junction of the culprit vein and thecal sac, often at the neck of the ipsilateral perineural cyst at the corresponding spinal level. Mamlouk et al. [25] described the technical procedural details and periprocedural considerations. Although treatment can be performed in isolation, we have had better results performing the procedure on the same day as the corresponding diagnostic dynamic CTM, as residual opacification of the perineural cyst and draining vein can aid procedural planning. As such, it is prudent to plan for potential treatment on the same day as dynamic CTM in patients with a high clinical suspicion of CVF.

Endovascular embolization—Brinjikji et al. [26, 27] and Borg et al. [28, 29] first described transvenous endovascular embolization of CVF in 2021 and subsequently showed the efficacy and safety of the procedure in a series of 40 patients. According to our algorithm, patients who fail to have a response to percutaneous fibrin glue embolization of CVF are evaluated for either endovascular embolization or surgical ligation. To perform endovascular embolization, the target paraspinal vein is catheterized, and a liquid embolic agent such as N-butyl cyanoacrylate (Johnson & Johnson) or Onyx (Medtronic) is used to occlude the CVF. Venous access is obtained through either the right common femoral vein or internal jugular vein to catheterize the azygos vein, vertebral vein, or ascending lumbar vein, depending on the level of the CVF in the spine. After the suspected paraspinal vein is catheterized, flat-panel cone-beam CT venography is acquired with images compared with previously performed dynamic CTM to confirm appropriate vein catheterization, followed by embolization (Fig. 7D). Additionally, anticipatory fiducial marker placement during dynamic CTM or CT fibrin glue injection can be useful to localize the appropriate spinal level for future procedures, minimizing procedure time.

Surgical ligation—Successful treatment of CVF via surgical ligation was first described by Schievink et al. [13]. According to our algorithm, CVFs that are refractory to percutaneous fibrin embolization and are otherwise unable to be targeted via endovascular techniques are considered for operative ligation. The surgical approach depends on the site and level of the fistula, but it often requires laminectomy and/or mesial facetectomy to gain access to the nerve root and adjacent veins (Fig. 9). Although the draining vein(s) may show hyperemia on direct inspection, the dominant culprit vein is often difficult to distinguish. Therefore, copious ligation and cauterization of venous vasculature surrounding the nerve root, with or without suture ligation of the nerve root itself, is often performed. During surgery, Valsalva maneuvers may assist in identifying intraoperative evidence of CSF leak.