The sella turcica is a spherical depression in the superior surface of the sphenoid bone. The sphenoid sinus is inferior and anterior to the sella turcica, the paired cavernous sinuses are lateral, the suprasellar cistern and its contents are superior, and the basilar artery and brainstem are posterior (Fig. 17.1). The pituitary gland, which weighs about 0.5 g in the adult, is the only structure of significance within the sella turcica. The dimensions of the pituitary gland are extremely variable, particularly its height. On average, it is 12 mm in width, 8 mm in anteroposterior diameter, and 3 to 8 mm in height. It achieves its greatest size in adolescent girls and pregnant women because of normal physiologic hypertrophy. At puberty, particularly in girls, the gland enlarges tremendously and may reach 10 mm in height. In boys the increase is more modest, with a height limit of about 8 mm. Even more marked changes occur in pregnancy. The gland increases progressively in size with gestation, reaching its maximum immediately postpartum. The maximum height may be as much as 12 mm, although usually not more than 10 mm. The gland height progressively decreases during the first 8 months after delivery, and the increased T1 signal intensity in the anterior gland normally seen in pregnancy progressively normalizes in the first 12 months after delivery. Termination of lactation has no effect on these changes.

The pituitary gland has anterior, intermediate, and posterior lobes. The intermediate lobe is vestigial in humans and serves no physiologic purpose, but it may be the site of small nonfunctional cysts (pars intermedia cysts). True vestigial pituitary gland cysts are found in 3% to 5% of autopsies (1). The anterior and posterior lobes are distinct organs functionally and embryologically. The anterior lobe, or adenohypophysis, originates from a superior invagination of the Rathke pouch from the fetal nasopharynx. Grossly, it is divided into three parts: the anterior lobe or pars distalis (the largest), the pars tuberalis surrounding the infundibulum, and the pars intermedia lying between the pars distalis and neurohypophysis. The anterior lobe loses all connections to the nasopharynx and develops into a true endocrine organ. The anterior lobe synthesizes many hormones under the influence of hypothalamic stimulatory and inhibitory controls. Adenohypophyseal cells have a definite topographic representation within the gland. Prolactin and growth hormone–secreting cells predominate laterally, whereas cells secreting corticotropin adrenocorticotropic hormone (ACTH), thyroid-stimulating hormone, and follicle-stimulating hormone/luteinizing hormone tend to be located centrally. This is a finding of some importance because the location of microadenomas parallels this distribution. The posterior lobe, or neurohypophysis, arises from the hypothalamus as a down growth into the sella turcica. The posterior lobe retains its connections to the hypothalamus throughout life via the pituitary stalk (the hypothalamohypophyseal tract) and functions only as a reservoir for the hormones, oxytocin and vasopressin (antidiuretic hormone). These hormones are synthesized in the hypothalamus and transported to the neurohypophysis down the axons of the pituitary stalk.

Although functionally distinct, the anterior and posterior lobes are anatomically closely related, apposed to one another within the confines of the sella turcica. The latter is the much smaller of the two lobes and occupies only 10% to 20% of the volume of the sella turcica, almost always in the midline and directly applied to the dorsum sella. The anterior lobe fills the anterior and central portions of the sella turcica and has two lateral wings that extend posteriorly, often to the dorsum sella. On rare occasions the lateral wings of the anterior lobe completely envelop the posterior lobe, so that the posterior lobe assumes a central location within the gland.

The pituitary gland receives its blood supply principally from the hypophyseal portal venous system. Small branches of the internal carotid artery—the inferior and superior hypophyseal arteries—also provide direct arterial supply. The inferior hypophyseal artery supplies the posterior lobe and the lateral surfaces of the anterior lobe, and the superior hypophyseal artery supplies the pituitary stalk and the superior surface of the gland. This combination of arterial and portal venous blood supply explains the differential rates of enhancement of the pituitary gland on dynamic scans after bolus administration of intravenous (i.v.) contrast agents (Fig. 17.2) (2). Within the gland, the vascular pattern is one of a loosely organized network of capillaries. The capillaries are lined by fenestrated endothelium, a characteristic feature of endocrine organs and one that is clearly visible in the pituitary gland. On the venous side, the gland is drained by the cavernous sinuses and afterward by the petrosal sinuses. There is preferential venous drainage of the gland to the ipsilateral cavernous sinus, but because there are extensive intercavernous venous connections anterior, posterior, and inferior to the gland without any valves to restrict directionality of flow, contralateral mixing of venous blood is possible. A variant large intercavernous sinus venous communication can occur along the floor of the sella, and should not be confused for an adenoma (Fig. 17.3A–C).

The anterior and posterior lobes of the pituitary gland are easily distinguished on magnetic resonance imaging (MRI) when typical, high-resolution studies of the pituitary region are obtained. With the exception of the neonate and in pregnancy, the anterior lobe is similar in signal intensity to cerebral white matter on all pulse sequences, whereas the posterior lobe is distinctly hyperintense on T1-weighted images (Figs. 17.1 and
17.2) (3). On administration of i.v. contrast, the anterior lobe, the posterior lobe, and the pituitary stalk all enhance intensely.

In neonates the anterior lobe has higher signal intensity on T1 weighting than in the adult (Fig. 17.4). At about 2 months of age, the intensity begins to diminish. By about 4 to 6 months of age, it resembles that of the adult (Fig. 17.4). In both pregnancy and in the neonate, lactotroph hypertrophy and increased protein synthesis in the pituitary gland have been proposed as explanations for the high signal.

The material within the posterior lobe responsible for the high signal intensity has long been debated (3,4). Various constituents of the neurohypophysis have been examined, including vasopressin, phospholipid, and neurophysin. The current thought is that the bright signal represents normal storage of vasopressin–neurophysin complex in the posterior lobe (4). One study demonstrated a strong correlation between vasopressin content in the posterior lobe and the signal intensity ratio between the posterior lobe and the pons. Furthermore, the hyperintensity
may not correspond precisely to the entire extent of the posterior lobe. Early postcontrast images in the sagittal plane show enhancement of an area in the posterior sella turcica larger than the posterior lobe hyperintensity visible on noncontrast images. This has led to the suggestion that the hyperintensity on noncontrast MR may only be located in the posterior portion of the posterior lobe, which contains more neurosecretory granules than the anterior part of the neurohypophysis (5). If the high signal on T1 MR is absent on high-resolution images, the list of possible causes includes central diabetes insipidus, persistent hypersecretion of ADH (as in dehydration), and a variety of stalk lesions interrupting the hypothalamic–neurohypophysial system (e.g., neoplasms like germinoma, inflammatory conditions, and in utero interruption with ectopic posterior lobe).

The sella turcica and the pituitary gland contained therein are covered by a fibrous sheet of dura, the diaphragma sella. The diaphragma sella is a normal dural structure composed of two layers forming a partial roof over the pituitary fossa. It extends from the tuberculum sella to the dorsum sella (6). The diaphragma sella has a small but variably sized opening (Fig. 17.5), the diaphragmatic hiatus, which allows passage of the pituitary stalk. Above the diaphragma sellae lies the suprasellar cistern. In the axial plane, the cistern resembles a six-pointed star: the interhemispheric fissure is anterior, the interpeduncular cistern is posterior, the paired sylvian fissures are anterolateral, and the ambient cisterns are posterolateral. The suprasellar cistern contains all the vascular anastomoses of the circle of Willis, the optic chiasm and nerves, the inferior hypothalamus, and the pituitary infundibulum with its investing venous plexus.

The pituitary stalk is the central landmark in the suprasellar cistern. It is approximately 2 mm thick, wider superiorly and tapering inferiorly. It slopes forward as it descends from the inferior hypothalamus, through the diaphragma sellae, to insert onto the superior surface of the pituitary gland at the junction of the anterior and posterior lobes. Above the infundibulum, the floor of the third ventricle tapers anteriorly to two pointed recesses: The infundibular recess converges toward the infundibulum, and the supraoptic (or chiasmatic) recess extends inferiorly to the chiasm (Fig. 17.1). In the setting of hydrocephalus, the infundibular recess may dilate and extend into the sella. A congenital dilation of the infundibular recess mimicking an infundibular stalk mass has been described (7). If thick sagittal or paramedian images are obtained, the stalk may appear thickened. On axial images, a thickened infundibulum with fluid attenuation centrally is seen. High-resolution sagittal and axial imaging will rule out a true mass. This normal variant can be distinguished from pathology by its smooth contour, the homogeneous enhancement and uniform thickness of the wall, the gradual downward tapering, and its communication with the third ventricle showing cerebrospinal fluid (CSF) attenuation on all pulse sequences. The optic chiasm is horizontally oriented and is immediately anterior to the pituitary stalk. When viewed from below, the chiasm resembles the letter “X.” Anteriorly, the chiasm divides into the optic nerves; these are closely applied to the inferior surface of the frontal
lobes. Posteriorly, the chiasm divides into the optic tracts. The paired optic tracts pass on either side of the midline stalk and paired mammillary bodies to enter the inferolateral aspect of the thalamus at the lateral geniculate bodies. The signal intensity of the optic chiasm, nerves, and tracts is similar to other compact white matter tracts on MR images.

The internal carotid artery (ICA) enters the suprasellar cistern on emerging from the cavernous sinus just medial to the anterior clinoid process. The carotid artery then gives off three medium-sized branches: ophthalmic anteriorly, posterior communicating, and anterior choroidal arteries posteriorly. Finally, the carotid artery terminally bifurcates into the anterior and middle cerebral arteries. The middle cerebral artery turns immediately laterally into the sylvian fissure, whereas the anterior cerebral artery turns medially and traverses the suprasellar cistern along the superior surface of the optic chiasm and nerves, and then anastomoses with the contralateral anterior cerebral artery via the anterior communicating artery. The internal carotid artery, middle cerebral artery, and anterior cerebral artery are easily visualized on MR because of their relatively large caliber and the absence of signal from the rapidly flowing blood in the vascular lumen (“flow void”) (Fig. 17.1). The smaller ophthalmic, anterior choroidal, and posterior communicating arteries are also often visible, particularly with newer high-resolution techniques and with MR angiography (MRA).

The paired cavernous sinuses are located immediately lateral to the sella turcica. These are venous structures that are functionally part of the dural venous system but anatomically are much more complex than any other dural sinus. They extend from the orbital fissures anteriorly to Meckel’s cave posteriorly and consist of an intricate network of small endothelial-lined spaces. They may communicate via a network of intercavernous connections through the sella turcica that are located anterior, inferior, and posterior to the pituitary gland (the “circular sinus”). Laterally, the cavernous sinuses are delineated by a well-defined dural reflection, separating the sinus from the middle cranial fossa. Medially, the dural reflection is weak, comprising only loose fibrous tissue or a thin, single layer of dura. Hence, it is not well visualized at imaging. It may be difficult to distinguish the exact boundary between cavernous sinus and the pituitary gland. On electron microscopy investigations, this medial dural wall is commonly incomplete even in normal subjects, so that normal pituitary tissue is found in the cavernous sinus in between 10% to 30% of subjects (8). The lateral wall of the cavernous sinus contains (from superior to inferior) cranial nerves III, IV, V₁, and V₂. These cranial nerves are typically recognizable on contrast-enhanced MR as nodular filling defects in the lateral cavernous sinus wall or adjacent to the cavernous carotid artery (Fig. 17.1). In fact, the absence of visualization of cranial nerve–related filling defects in the enhanced cavernous sinus should prompt suspicion of a lesion in the cavernous sinus. Within the sinus the ICA is easily recognized as the largest structure contained therein (Fig. 17.1). Cranial nerve VI is also within the sinus, directly beneath and intimately associated with the internal carotid artery. The appearance of the venous spaces is heterogeneous. Much of the sinus is gray, whereas some parts display a more typical flow void. Presumably, the latter is from larger channels with higher rates of blood flow and the former
from smaller channels with sluggish flow. The entire cavernous sinus, with the exception of the ICA and the cranial nerves, enhances intensely with i.v. contrast administration.

Meckel’s cave is closely related to but anatomically distinct from the cavernous sinus. It is a dural invagination that arises from the vertical surface of the anterior wall of the posterior cranial fossa. The orifice through which it communicates with the basal cisterns is inferior to the petroclinoid ligament. It is immediately lateral to the posterior third of the cavernous sinus and houses the trigeminal ganglion. Cranial nerve V₃ is identified passing inferolaterally from Meckel’s cave through the foramen ovale directly beneath and slightly lateral to the Meckel’s cave (Fig. 17.1).

Suspected abnormalities of the pituitary gland and sella turcica are among the most frequent indications for MR. An effective MR protocol that considers image contrast and spatial resolution requirements yet optimizes scanning time is necessary. As in all MR protocols, the overall objective is to obtain the most spatially detailed images in the appropriate plane, best possible signal-to-noise ratio, and highest image contrast within the shortest period of time. The most important goal in imaging a patient suspected of harboring pituitary or suprasellar pathology is most often the identification of the pituitary gland itself and the separation of it, if possible, from the lesion. Therefore, it is essential to use a high-resolution MR protocol that clearly demonstrates the anatomy of this specialized region.

It is generally accepted that the coronal plane is the most useful plane or section for imaging the pituitary gland. This plane allows visualization of the pituitary gland free of partial averaging effects from the carotid arteries, sphenoid sinus, and suprasellar cistern. Coronal images are virtually always considered in conjunction
with an imaging series in the sagittal plane, primarily for display of midline structures. Practically speaking, most scanning protocols start with a sagittal sequence for localization; for patients imaged for suspected pituitary pathology, this initial sagittal sequence should be performed with a high-resolution technique.

Exact specifications of technique are not provided because the constantly improving capabilities of modern scanners require that technique is continually revised. However, high spatial detail is critical and should be achieved through the use of thin (3 mm or smaller) slices, a fine matrix size, and a relatively small field of view. Requirements for spatial detail need to be balanced against signal-to-noise ratio and imaging time. We prefer to limit individual sequences to 5 to 6 minutes, and certainly no longer than 10 minutes.

The pulse sequence for best tissue contrast and anatomic display in this region is generally accepted to be T1-weighted imaging, although there may still be some controversy as to what sequences are needed in addition to this. Spin echo (SE) methods were the first to gain widespread popularity and continue to be the most widely used. Several groups have shown that short-repetition-time, short-echo-time SE images (i.e., T1 weighted) generate very good contrast for visualizing pituitary pathology (9). Even though high-resolution T1-weighted images are now standard for pituitary pathology, they are not artifact free. An unusual artifact in the sella manifesting as focal T1 hyperintensity has been described in 14% of patients (10). Its presence could potentially obscure a lesion or could be mistaken for substances resulting in T1 shortening such as hemorrhage or protein. The artifact is secondary to focal susceptibility differences among the air-filled sphenoid sinus, bone, and the pituitary fossa soft tissue and is consistently related to the junction between the sphenoidal septum and sellar floor. Hyperintensity along the floor of the pituitary fossa posteriorly may also be due to the normal posterior lobe signal, a useful landmark to seek out for signs of lateral displacement by microadenomas. Compared with T1-weighted images, long–repetition time, long–echo time SE (i.e., T2-weighted) images have been less successful in demonstrating pituitary lesions, in particular small adenomas (9,11). However, there are occasional cases in which only the T2-weighted image has been positive (Fig. 17.6). Because fast spin-echo (FSE) methods are readily available and do not incur the time penalty of conventional T2-weighted SE sequences, T2-weighted FSE imaging is sometimes used as a supplementary sequence for pituitary imaging.

A variety of other pulse sequences have met with varying degrees of success, including inversion recovery and gradient-echo (GE) methods. Multislice inversion recovery sequences have not been widely available and have been infrequently used. Two-dimensional (2D) Fourier transform GE images have been unsuccessful primarily because of the marked susceptibility artifacts generated from the interface between the skull base and air-containing sphenoid sinus. On the other hand, 3D Fourier transform GE images have achieved greater success; these have been found to provide results comparable with T1-weighted SE methods (12). They have the advantage of thinner slices compared with conventional 2D Fourier transform SE methods, yielding better spatial resolution. In addition, the voxels are more nearly isotropic, and truly contiguous slices are possible. However, they are prone to motion and truncation artifacts in the two dimensions that are phase encoded. Patronas et al. (13) reported improved visualization of ACTH-secreting adenomas with the use of spoiled gradient recalled acquisition over SE T1 imaging.

Paramagnetic contrast-enhanced images are widely used and can be very helpful in this region. Although most adenomas are visible without the injection of i.v. contrast, several studies have shown that small adenomas may become visible only after contrast injection (14). Many radiologists, including ourselves, have reduced the amount of contrast used for the investigation of pituitary adenomas to half the usual dose (to 0.05 mmol/kg). It has been demonstrated that half-dose studies have comparable results to full-dose studies in terms of pituitary gland opacification and microadenoma detection (15). Although full-dose and double-dose studies have demonstrated a greater absolute signal intensity difference between gland and lesion, the ratio of gland to lesion signal is similar when comparing half-dose, full-dose, and double-dose results. The greater absolute gland to lesion signal differences may improve lesion detection and characterization, but whether this is clinically relevant is unknown. It can be argued, on the other hand, that the MR workup for hyperprolactinemia, the most common clinical impetus for imaging the pituitary, is mainly performed for the purpose of excluding disease other than pituitary microadenoma because at least in some centers the treatment of choice is medical rather than surgical. This would make the identification of the microadenoma less important, hence reducing the need for i.v. contrast for the pituitary gland itself. On the other hand, many institutions include post–gadolinium enhanced images of the brain, thereby
making it more important to use the full, standard dose of contrast agent.

Because of the differential rates of contrast enhancement between adenomas and normal pituitary gland, lesion conspicuity can be optimized by scanning immediately after administration of the contrast agent. Immediately after contrast injection, most adenomas will appear as relatively nonenhancing (dark) lesions within an intensely enhancing pituitary gland. However, soon afterward this image contrast begins to dissipate, and the adenoma may no longer be detectable. The development of faster and faster imaging sequences has generated some renewed interest in “dynamic imaging” of the pituitary gland. Dynamic imaging sequences may have temporal resolution as fast as 1 to 2 seconds per set of images, although protocols range from images every second to images every minute (16,17,18). Our experience with dynamic scanning has shown that the contrast enhancement behavior of microadenomas is inconsistent from case to case. Some microadenomas display the most lesion-to-gland contrast on unenhanced scans. With these, the image contrast begins to diminish the moment the contrast-enhancing agent arrives in the pituitary gland. In about 20% of cases, the maximum image contrast between normal pituitary tissue and microadenomas is attained about 30 to 50 seconds after the bolus injection of i.v. contrast, whereas others have the best image contrast 1 to 2 minutes after contrast agent injection (16). Thus, the best images for microadenoma detection vary from the unenhanced images, the 30- to 50-second dynamic-enhanced images, to the conventional enhanced images. When we use dynamic imaging techniques, our preference for dynamic scanning is a modified T1-weighted FSE method with 10-second temporal resolution (16). We found an incremental yield in microadenoma detection of about 10% above and beyond that detected on a combination of plain and conventional contrast-enhanced MR (Fig. 17.7). Gao et al. (17) found that adding a sagittal dynamic study in addition to the traditional coronal study increased lesion detection, with an overall detection rate of 89%. Faster dynamic sequences with temporal resolution of 5 to 10 seconds have demonstrated an early component of enhancement in microadenomas not visible on sequences with slower resolution. These fast images have shown that some microadenomas have subtle enhancement early, at about the same time as the posterior lobe and long before the anterior lobe. This confirms earlier computed tomography (CT) work of Bonneville and supports the conclusion that these adenomas have a direct arterial supply from branches of the carotid artery and therefore would not be subject to the normal regulatory influences of the hypothalamus via the portal venous system (18). Dynamic imaging may also be helpful in characterizing lesions adjacent to the pituitary gland, such as meningiomas or cavernous malformations, due to the differential enhancement patterns.

Delayed scans (i.e., longer than 30 minutes after contrast injection) occasionally may demonstrate a reversal of this image contrast. As the contrast agent diffuses into the adenoma, the adenoma becomes hyperintense and the intensity of the normal pituitary gland fades, allowing the adenoma to stand out as a hyperintense focus. Unfortunately, delayed scans are not practical because they require that the patient either wait in the scanner or return for reexamination 30 to 60 minutes later. Thus, they are infrequently used.

Our current practice is to perform unenhanced T1-weighted SE sagittal and coronal sequences. If these are positive for an intrapituitary lesion, the examination stops. If they are negative or equivocal, a bolus of i.v. contrast is given synchronously with the beginning of a dynamic T1-weighted FSE sequence. At the completion of the dynamic study, we immediately proceed to a conventional T1-weighted SE sequence. If these are negative, equivocal, or discordant with one another or if there is evidence of a suprasellar or hypothalamic lesion, T2-weighted images are obtained; otherwise, the examination ends. In a less monitored situation, after high-resolution T1-weighted imaging through the pituitary fossa, the remainder of the brain is always scanned using T2-weighted imaging, and finally, postcontrast
T1-weighted coronal and sagittal images of the pituitary are obtained.

Magnetization transfer techniques provide tissue contrast dependent mainly on the concentration of macromolecules, quantified by the magnetization transfer rate (MTR). The MTR of the normal pituitary varies with age. The MTR has been shown to be higher in prolactin-secreting or growth hormone–secreting adenomas compared to normals and lower in nonsecreting adenomas compared to normals (19).

The ultimate clinical role of 3-T pituitary and sellar imaging continues to evolve and further clarification of the use of newer sequences is required (Fig. 17.8). The images offer improved image quality and spatial resolution, which is important in conditions with subtle differences between normal and abnormal tissues, although there can be more magnetic susceptibility–related signal loss in proximity to air-filled sphenoid sinus. One study demonstrated better preoperative localization of pituitary of pituitary macroadenomas in Cushing’s disease with 3-T compared to 1.5-T MRI (20). Greater clarity of the intracavernous cranial nerves, assessment of the pituitary stalk, and subtle changes in gland volume have been demonstrated (21,22). Greater accuracy at detecting cavernous sinus invasion compared to 1.5-T imaging has been reported. Improved image quality with better definition of the borders of sellar lesions and delineation of cranial nerves was achieved with a 3D GE sequence compared to an SE sequence on contrast-enhanced T1-weighted image at 3 T (23).