Practice Essentials
Cerebrospinal fluid (CSF) leak may occur from the nose (rhinorrhea), from the external auditory canal (otorrhea), or from a traumatic or operative defect in the skull or spine. [1] CSF leaks occur when there is a tear or hole in the dura mater, which is the outermost layer of the meninges that protects the central nervous system. This leak of fluid can cause low-pressure headaches, neck pain, ringing in the ear, and, occasionally, loss of smell or taste. According to Ommaya et al, 80% of CSF leaks are due to nonsurgical trauma, 16% are iatrogenic, and 4% are spontaneous. The most common locations of a CSF leak are the upper thoracic level (T1-T6) and the lower thoracic level (T7-T12). [2, 3, 4, 5, 6]
The fluid leak is a result of meningeal dural and arachnoid laceration with fistula formation. Blunt trauma is the most common cause. Traumatic CSF leak is reported in approximately 10-30% of skull base fractures in adults. More than half of these present within 48 hours. The most common fracture sites leading to CSF leaks are the frontal sinus (30.8%), sphenoid sinus (11.4-30.8%), ethmoid (15.4-19.1%), cribriform plate (7.7%), frontoethmoid (7.7%), and sphenoethmoid (7.7%). [7]
(See the images below.)
Computed Tomography
CT findings associated with cerebrospinal fluid leaks include fractures or other bone defects; meningocele; focal fluid accumulation in the ethmoid air cells; frontal, sphenoid, or maxillary sinuses or mastoid air cells; and, sometimes, pneumocephalus.
(See the images below.)




CT cranial cisternography is performed with injection of 5-7 mL of nonionic myelographic contrast medium into the lumbar subarachnoid space. The patient is maintained in the prone position until a CT scan is performed. Ideally, the contrast medium is concentrated in the intracranial anterior and posterior skull base regions under fluoroscopic guidance by tilting the prone patient head downward on a fluoroscopic tilt table. Alternatively, with the patient lying prone on a stretcher, the patient's hips can be raised above the level of the head for 1-2 minutes to concentrate the contrast medium over the anterior and posterior regions of the skull base. Coronal CT images of 2-3 mm thickness are then obtained through the face and cranium, including all of the paranasal sinuses and the mastoid air cells.
CT myelography is used in the detection of spinal CSF leak. Slow CSF leaks may be detected by postmyelogram CT scan in which there is a time delay between the contrast medium intrathecal injection in the fluoroscopic room and subsequent transfer to the CT scan room. Fast CSF leaks have rapid contrast diffusion and may not be localized to a 2-vertebral segment of the spinal canal (suitable for local treatment by extradural blood patch or alternate therapy) by routine postmyelogram CT spine scan. A high percentage of fast leaks have spinal extradural fluid collections on preliminary MRI spine scans. Dynamic CT myelography is recommended in these patients, with the injection of the iodinated contrast medium intrathecal on the CT scan table with immediate spine CT scan. [26, 27]
CT cisternographic findings in CSF leak include the concentration of contrast medium in portions of a paranasal sinus or within ethmoid or mastoid air cells. Occasionally, a stream of contrast medium is demonstrated at the fistula site.
Digital subtraction radiographic cisternography can be similarly performed with a spinal subarachnoid injection of nonionic iodinated contrast medium. The images may demonstrate a CSF fistula, but this technique is used less frequently than other cisternographic methods.
Degree of confidence
The incidence of CSF fistula detection varies from 22 to 100% in clinical studies. The fistula detection rate is lowest for intermittent CSF leaks. The accuracy of active fistula detection with CT cisternography is 65-85%. In one study of 45 patients, CT of the skull and facial bones with high-resolution, thin-section axial and coronal images had an accuracy of 92%, a sensitivity of 92%, and a specificity of 100% in depicting the presence or absence of CSF fistula. [28] Contemporary computer-reconstructed coronal images are usually of diagnostic quality, and direct CT coronal images may not be necessary.
Magnetic Resonance Imaging
Brain tissue herniation is best seen on MRI. Herniation of the inferior frontal gyrus may occur in frontal head injuries or in ethmoid developmental defects of the cribriform plate. Temporal lobe gyral herniation may occur through a petrous temporal bone tegmen tympani defect. [29, 30] (See images below.)





In spontaneous intracranial hypotension syndrome (SIHS), brain MRI shows thickening and contrast enhancement in the cranial pachymeninges. Subdural hygroma or hematoma on the cerebral convexities is common. The brain is noted to sink downward in the cranium with development of a pseudo-Chiari I malformation. The cerebral dural venous sinuses may be engorged. The cerebral ventricles may be reduced in size, and the pituitary gland may appear enlarged. There may be apparent downward displacement of the optic chiasm. The upper cervical epidural veins are congested. All of these changes are reversible with ablation of the cause of CSF leak, which is usually in the spine.
Spinal MRI in patients with SIHS may show some irregularity of the thecal sac due to partial dural collapse. Extradural fluid collections are common in spinal CSF leak. Intense extradural contrast enhancement is noted in congested epidural veins. One or more CSF fistulas may originate from spinal nerve root sleeves in the case of spontaneous spinal CSF leak.
The localization of one or multiple leaks can make possible and facilitate therapeutic CT-guided epidural blood patching. [31]
A variety of cisternographic studies may be necessary to localize some spinal CSF fistulas. Spinal MRI findings are also potentially reversible after successful ablation of a CSF fistula.
MR cisternography and myelography
MR cisternography and myelography can accurately localize CSF leaks in the cranium and spine. [32, 33, 34, 35] This technique is based on the intrinsic T2 contrast between CSF and adjacent structures. A positive diagnosis of CSF fistula is made by finding direct continuity of the CSF fistula with the subarachnoid space. Coronal and sagittal imaging is necessary.
(See images below.)




The high T2 signal from CSF fistula may be difficult to differentiate from that of sinusitis on axial images. A high rate of fistula detection may be possible with imaging in the prone position, but this may be uncomfortable for the patient. Therefore, imaging is usually done with the patient in the supine position. Rapid echo-planar imaging with the patient in the prone position and performing a Valsalva maneuver may allow for limited coronal imaging and increase the accuracy of MR cisternography.
MR cisternography is performed with heavily T2-weighted, fast spin-echo, fat-saturated sequences with thin sections and minimal or no gap. Typical imaging parameters include a repetition time of 10,000 ms, an effective echo time of 200 ms, 4 signals acquired, an echo train length of 16, a matrix of 512 X 192, no phase-wrap option, 3-mm sections interleaved contiguously (0-mm gap), and a 16-cm field of view.
A short repetition time can be used to achieve a result similar to that of the technique above, with slightly faster imaging times. The gray scale is reversed for optimal viewing. With one method, the average total time for coronal and sagittal imaging is 48 minutes. [36] Most MRI machines offer fat suppression and image gray-scale reversal. Additional hardware or software is not required to perform MR myelography or cisternography.
MR cisternography may demonstrate inactive CSF fistulas. MR T2 myelography may demonstrate spinal CSF fistulas (see the images below). The intrathecal injection of 0.5 mL of gadopentetate dimeglumine diluted in 3-5 mL of CSF for MR cisternography has been reported to have high sensitivity and specificity for detection of active CSF fistula, exceeding the rate of fistula demonstration by CT, nuclear medicine, or noncontrasted MR cisternography.
Small series of patients had no apparent adverse effect from the gadolinium contrast medium. [34, 37] Gadolinium-based contrast media are approved for intravenous injection for MRI but have not been approved for intrathecal use in humans by the Food and Drug Administration (FDA). The intrathecal injection of gadolinium-based contrast media has been shown in several off-label studies to be effective and safe in selected patients in whom other cisternographic or myelographic studies have failed to demonstrate the CSF leak site. [38, 39] Severe brain injury has been reported in a patient who received, erroneously, 30 times the intended dose of gadolinium in an MRI myelogram. [40]
In a small series of patients with spontaneous intracranial hypotension (SIH), instillation of preservative-free normal saline into the thecal sac followed by intrathecal gadolinium infusion was found to be a safe technique that increased the detection of a CSF leak on MR myelography images. [41]



Gadolinium-based contrast agents have been linked to the development of nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD). NSF/NFD has occurred in patients with moderate to end-stage renal disease who have been given a gadolinium-based contrast agent to enhance MRI or MRA scans. NSF/NFD is a debilitating and sometimes fatal disease. Characteristics include red or dark patches on the skin; burning, itching, swelling, hardening, and tightening of the skin; yellow spots on the sclera of the eyes; joint stiffness with difficulty moving or straightening the arms, hands, legs, or feet; pain deep in the hips or ribs; and muscle weakness.
Nuclear Imaging
Radionuclide cisternography is performed by administering a lumbar subarachnoid intrathecal injection of Indium-111 (111In) diethylenetriamine pentaacetic acid (DTPA) in a 500 µCi dose. In-111 has minimal background activity and does not accumulate in the brain. Technetium as 99mTc DTPA is a less frequently used isotope. [42] The sensitivity for CSF leaks is in the range of 50-100%. The specificity is almost 100% for contemporary radionuclide cisternography.
(See images below.)



Head images are acquired 2, 6, 12, and 24 hours after injection of the isotope. Follow-up 48- or 72-hour scans are possible with 111In and may be useful in the detection of intermittent CSF fluid leaks.
The entire spine is scanned up to 24 hours in cases of spontaneous intracranial hypotension, spinal trauma, or postoperative CSF leaks. [43] Cotton pledgets labeled for the placement site are positioned in the nose before the lumbar subarachnoid space injection of the isotope. Pledgets are placed close to the cribriform plate, in the middle meatus, and in the sphenoethmoidal recess of the right and left nasal cavities. A control pledget for lacrimal secretions is placed under one inferior nasal turbinate.
Pledgets are scanned in a glass tube at intervals of 2-24 hours, with the highest count rate indicating a possible leak site. Alternatively, radioactivity of the nasal pledgets is compared to that of known plasma radioactivity.
For otorrhea, 1 cotton pledget is placed in each external auditory canal.
-
Lateral 24-hour cranial scintigraphic image from a nuclear medicine cisternographic study in a patient with clinically evident right-sided cerebrospinal fluid rhinorrhea. Image demonstrates increased tracer accumulation in the nasal region (arrow).
-
Anterior 48-hour scintigraphic image demonstrates tracer accumulation in the right nasal region. Imaging findings were correlated with both the clinical findings and nasal pledget counts obtained as part of this study.
-
Acute posttraumatic cerebrospinal fluid rhinorrhea. This coronal magnetic resonance cisternogram demonstrates a left-sided cerebrospinal fluid leak through the cribriform plate (small arrows), which was clinically suspected. The image also shows a right-sided meningocele (large arrow) protruding through the cribriform plate, which was not suspected but was surgically repaired at the same time as the left cribriform cerebrospinal fluid leak site.
-
This patient presented with a spontaneous onset of cerebrospinal fluid rhinorrhea 10 years after a head injury. This coronal CT cisternogram was obtained after an intrathecal injection of contrast material (Omnipaque 300, 8 mL) into the lumbar thecal sac and subsequent positioning of the contrast agent in the head. The image demonstrates dense contrast medium layering in the empty sella and contained within the meningocele (arrow).
-
Axial CT image was obtained with the patient in the supine position. Contrast medium has drained out of the meningocele, but a small amount remains in the sphenoid sinus around the meningocele.
-
Magnetic resonance cisternogram with cerebrospinal fluid rhinorrhea demonstrates a meningocele extending into the left lateral recess of the sphenoid sinus (arrows).
-
Axial magnetic resonance cisternogram demonstrates the connection of the meningocele to the middle cranial fossa (arrows). Fluid contained in the meningocele and leaked fluid in the sphenoid sinus outline the meningocele membrane.
-
Sagittal magnetic resonance cisternogram demonstrates the connection of the meningocele to the middle cranial fossa; this finding facilitated surgical planning.
-
Fast spin-echo T2-weighted coronal image of a patient with a spontaneous onset of cerebrospinal fluid rhinorrhea demonstrates an empty-sella configuration.
-
Axial CT image demonstrates pneumocephalus in association with the spontaneous cerebrospinal fluid rhinorrhea and a septal bone defect in the left posterior ethmoid air cell.
-
Coronal CT image of the temporal bone demonstrates a bone defect (small arrows) in the tegmen tympani with a protruding soft-tissue meningoencephalocele (large arrows). This patient had cerebrospinal fluid otorrhea after mastoidectomy.
-
Coronal fast spin-echo T2-weighted image demonstrates herniation of meninges and brain tissue (arrows) with adjacent cerebrospinal fluid into the postmastoidectomy tegmen tympani defect. This finding is consistent with a meningoencephalocele of the temporal bone.
-
Artist's rendering of a tegmen tympani bone defect with a herniated meningoencephalocele.
-
Sagittal magnetic resonance myelogram demonstrates a traumatic cerebrospinal fluid leak (small arrows) with disruption of the ligamentum flavum posteriorly (large arrow).
-
Magnetic resonance myelogram in a patient with a brachial plexus injury and pseudomeningoceles (arrows).
-
Magnetic resonance myelogram demonstrates pseudomeningoceles secondary to a stretch injury of the lumbosacral nerve roots.
-
Spontaneous intracranial hypotension syndrome in a patient with chronic headaches, which began after lumbar puncture. Axial fast spin-echo T2-weighted MRI demonstrates widened extra-axial fluid spaces but no focal extra-axial fluid collection.
-
Coronal fast spin-echo T2-weighted MRI.
-
Gadolinium-enhanced T1-weighted axial MRI shows diffuse moderate dural thickening with contrast enhancement.
-
Gadolinium-enhanced, coronal, T1-weighted MRI shows dural and tentorial thickening with contrast enhancement.
-
Nuclear cisternogram obtained at 24 hours demonstrates diffuse epidural accumulation of the tracer in the midlumbar region. This finding is suggestive of a site of cerebrospinal fluid leak.
-
Gadolinium-enhanced, T1-weighted axial MRI obtained 2 weeks after a 7-mL extradural blood patch was applied to the midlumbar region. This image shows complete resolution of the previous dural thickening and contrast enhancement. The patient's severe postural headaches were markedly decreased in intensity.
-
Gadolinium-enhanced, coronal, T1-weighted MRI.