Optic Pathway Gliomas
The most important primary brain abnormality in NF1 is the OPG (
25,
55,
56). OPG is seen in 15% to 20% of individuals with NF1; almost all present before age 6 years. Less than one-half of these tumors cause vision loss; very few cause precocious puberty (
57). Chiasmatic and retrochiasmatic lesions carry the highest risk for progression and visual loss (
56,
58,
59); however, imaging features that predict development of symptomatic OPG have not been identified (
60). Early identification of OPG by “screening MRI exams” (i.e., MRI exams in asymptomatic NF1 children) may lead to improved visual outcomes (
58).
The tumor can be isolated to a single optic nerve or can involve both optic nerves; the chiasm and the optic tracts can also be involved (
Figs. 6-1,
6-2,
6-3 and
6-4). Primary involvement of the optic nerve is more common in the NF1 population, while sporadic OPG (in patients without NF1) more commonly involves the chiasm and optic radiations (
61). Irrespective of the distribution, the clinical course of NF1-associated OPG is more indolent than in the sporadic OPG (
61,
62).
Optic pathway tumors (and other tumors that arise commonly arise in the diencephalon, near the floor of the third ventricle in NF1 patients) are most commonly of low histological grade; most are pilocytic astrocytomas. In some NF1 patients, these tumors can spontaneously involute over time (
Fig. 6-5) (
63,
64,
65). Therefore, aggressive therapy is usually not indicated unless symptoms progress or radiologic evidence of progression is noted. Uncommonly, optic pathway tumors can be highly malignant, demanding a very aggressive treatment approach (
59).
Tumors of the optic nerve infiltrate and expand the nerve, producing a fusiform mass. Two architectural forms can be distinguished. The
intraneural form is a diffuse expansion of the optic nerve itself without
significant subarachnoid tumor (
Fig. 6-2C). The
perineural form is tumor infiltration of the subarachnoid space (astrocytic proliferation), creating a rim of tumor around a relatively unaffected nerve; at times, both forms can coexist (
Fig. 6-3B) (
66,
67). Postcontrast MR using fat suppression will distinguish the two architectural forms of optic nerve gliomas. When the nerve is diffusely infiltrated, enhancing tumor fills the optic nerve sheath, whereas infiltration of the subarachnoid spaces shows a rim of enhancing tumor around a minimally enhancing optic nerve. Marginal cases (characterized by development of mild thickening of the optic nerves or by transient or mild enhancement of the optic nerves) are commonly observed; it is not clear whether these truly warrant a diagnosis of OPG (
60). Finally, optic nerve tortuosity (without abnormal thickening or enhancement) is commonly observed in NF1 and should not be mistaken for an optic nerve glioma (
60,
68).
Although tumors of the intraorbital optic nerves can be evaluated with CT, MR is the preferred imaging modality for evaluating the entire CNS in NF1 due to the increased information and lack of exposure of the child to ionizing radiation. Assessment of the precise diameter of the intraorbital optic nerves may be hampered by chemical shift artifact resulting from the location of the nerve adjacent to orbital fat, but use of fat-suppression techniques eliminates the chemical shift artifact and allows the nerve to be excellently evaluated, along with the intracranial optic nerves, optic chiasm, optic tract, and optic radiations. T1 and fatsuppressed T2-weighted axial and coronal sequences, using less than or equal to 3 mm slice thickness, should be obtained through the globes, optic nerves, and optic chiasm and followed by fat-suppressed contrastenhanced T1 axial and coronal images. A routine brain scan should follow to look for involvement of the optic tracts and other areas of the brain. Fluid-attenuated inversion recovery (FLAIR) and T2-weighted images show involvement of the optic tracts as high signal intensity extending posteriorly, often to the level of the lateral geniculate bodies; the internal capsules and adjacent basal ganglia are commonly infiltrated (
Figs. 6-1 and
6-2). Further extension beyond the optic pathways can also be identified in some patients: superiorly into the hypothalamus, fornices, and
septum pellucidum; laterally into the temporal lobes; posteriorly into the optic radiations; and inferiorly into the cerebral peduncles and the brainstem (
Fig. 6-2). Rarely, tumor can extend into the lateral ventricles.
Care should be taken not to mistake enlarged subarachnoid spaces within the optic nerve sheaths (the result of ectasia of the dura surrounding the optic nerves) for optic nerve tumors. The enlarged perioptic spaces have signal intensity identical to cerebrospinal fluid (CSF) and do not enhance after contrast administration. Care should also be taken not to confuse extension of OPG into the optic tracts (and beyond) with the characteristic T2 bright foci encountered in pediatric patients with NF1 (see next section). Differentiation of the two pathologies can be very difficult; features of OPG infiltration include contiguity, mass effect, low T1 signal on precontrast images, enhancement after contrast administration, and significant elevations of choline (at MR spectroscopy) (
Fig. 6-4) (
69).
Other Gliomas and Tumor-Like Conditions
Astrocytomas are more common in NF1 than in the general population (
25,
48,
55,
70); they can arise anywhere in the CNS. Pilocytic astrocytomas (
Figs. 6-6 and
6-7) are most common, but other low-grade (
71) and higher-grade tumors (
70,
72) also occur. Almost any part of the CNS can be affected; outside of the optic pathway/diencephalon, the most common site is the brainstem (
Fig. 6-7) (
73). It is of interest to note that the brainstem tumors occurring in NF1 seem to differ biologically from brainstem tumors in other patients. In contrast to the predominance of pontine tumors in the general population, the medulla is the most common site of brainstem tumor origin in NF1, followed by the mesencephalon (particularly the periaqueductal region/tectum) and the pons. This difference in location may partly explain the indolent course and vastly better long-term outcome of most brainstem tumors in NF1 (
73,
74,
75,
76). Many tumors do not require therapeutic intervention (
52); some may regress spontaneously (
77). Thus, management of brainstem tumors in children with NF1 is generally not aggressive unless clinical progression is seen—isolated radiologic progression is considered less serious. Other common locations for astrocytomas in children with NF1 are the cerebellum (
Fig. 6-6), the corpus callosum, and the cerebral hemispheres (
Fig. 6-8) (
48,
55,
78).
Chronic, nonprogressive enlargement of the pons, medulla, and middle cerebellar peduncles is occasionally observed in children with NF1 in association with, and likely the result of, extensive myelin
vacuolation (
Fig. 6-9). This hamartomatous enlargement should not be misdiagnosed as a brainstem glioma; differentiating features from infiltrating gliomas include normal or nearly normal signal in T1-weighted images, only modest (and usually heterogeneous) signal increase on T2 images, absence of contrast enhancement, and lack of progression on follow-up scans. The T2 signal abnormalities partly regress during adolescence, though the brainstem remains enlarged. For more discussion of brainstem tumors, see
Chapter 7.
Hamartomatous enlargement of the hypothalamus is occasionally seen in children with NF1 (
Fig. 6-10), usually not associated with precocious puberty or gelastic seizures (
79). The lesion typically lies within the walls of the hypothalamus in the lower third ventricle (intrathalamic/sessile configuration). Like hypothalamic hamartomas in children without NF1, they are nearly isointense to gray matter on T1- and T2-weighted images and do not enhance (
80); they should not be confused for a chiasmatic or hypothalamic glioma.
In a small but significant number of patients with NF1, hydrocephalus is present (
81,
82,
83). The site of obstruction of CSF flow is usually the aqueduct of Sylvius, resulting from benign aqueductal stenosis, from extensive vacuolation of the tissues surrounding the aqueduct (see next section), or from tumors (astrocytomas or hamartomas) of the adjacent tectum or tegmentum of the mesencephalon. The diagnosis of a tectal glioma is relatively simple using MR. In contrast to benign aqueductal stenosis (in which the proximal aqueduct is dilated and the tectum displaced superiorly and thinned, see
Chapter 8), gliomas enlarge the tectum and narrow (or completely obliterate) the aqueduct.
Occasionally, the tectum may appear short and thick in aqueductal stenosis because of mass effect on the rostral aspect of the tectum by a dilated suprapineal recess (see
Chapter 8). In these cases, the patient should be reevaluated after ventricular decompression.
White Matter Abnormalities and Characteristic Hyperintense T2 Lesions
White matter abnormalities are evident in many NF1 children. Many patients have an increased volume of cerebral white matter, typically associated with a large midsagittal corpus callosum area (
Fig. 6-11) (
84,
85); these findings are more pronounced in patients with macrocephaly (˜40% of those with NF1 (
86)) and are associated with a lower overall IQ (
87). Advanced MRI diffusion analysis of the white matter of NF1 patients reveals a decreased fractional anisotropy (FA) and an increased apparent diffusion coefficient (ADC). Radial diffusivity is disproportionately increased, suggesting that looser packing of axons, with or without myelination changes, may contribute to the increased volume of cerebral white matter (
88). Abnormalities of apoptosis (leading to redundancy of fibers connecting the cerebral hemispheres) have also been postulated as an explanation of the increased white matter volume (
87).
MR imaging often reveals characteristic hyperintense T2 lesions in pediatric patients with NF1, located in the brainstem, cerebellar white matter, basal ganglia (especially the globus pallidus), thalamus, internal capsule, corpus callosum, and, occasionally, corona radiata (
Figs. 6-12 and
6-13) (
36,
89,
90) but not in the centrum semiovale or the subcortical white matter. These lesions are characteristically multiple; they have little or no mass effect (unless extensive confluence is evident, in which case mild mass effect may be detected,
Fig. 6-9) and do not elicit vasogenic edema; their T1 signal appears iso- or slightly hypointense compared to uninvolved white matter; and they do not enhance after intravenous administration of paramagnetic contrast. Intermediate echo (TE = 144 ms) proton MR spectroscopy shows nearly normal
N-acetylaspartate (NAA)/creatine (Cr) and moderately increased Cho/Cr (
69). These lesions are rarely identified in the first 2 years of life but develop/become visible from late infancy to about age 12 years (
Fig. 6-12) (
89); new lesions almost never develop after age 15 years. The lesions start to regress starting late in the first decade of life (
90,
91) and slowly disappear; they are almost never seen in patients over the age of 20 years (
Fig. 6-13). In the peak age group (6-12 years), they are seen in up to 90% of NF1 children (
39,
92,
93). A pathologic analysis has shown that these regions contain areas of myelin vacuolation, where the layers of myelin become separated as they spiral around the axon (
94). As might be expected, water diffusivity is increased in these areas of T2 hyperintensity to a greater degree than in other areas of the brain (
95); anisotropy is normal, but transverse eigenvalues are increased (
96). Multishell diffusion MRI studies have shown that the water contained in the T2 hyperintense lesions behaves as intracellular water that has obtained extracellular-like properties, thus supporting the intramyelinic edema appearance of vacuolation (
97).
The discoveries that the oligodendrocyte myelin glycoprotein gene is imbedded within the
NF1 gene (
19) and that its protein product (neurofibromin) is required for Schwann cell myelination (
18) further support dysgenesis of myelin as the cause of the white matter abnormalities in NF1. The vacuolization appears to develop, therefore, in regions where the myelin is inherently abnormal; presumably, the return of normal T2/FLAIR signal on subsequent exams reflects myelin
repair or remyelination (
98). The increased diffusivity of white matter (even normal-appearing white matter) in NF1 (
99,
100) further supports the concept of abnormal myelin as an underlying cause.
In view of the frequency with which the characteristic T2 hyperintense foci are encountered in children with NF1, neither close imaging follow-up nor biopsy is indicated when they occur in the characteristic locations, are not significantly hypointense on T1-weighted images, and lack associated mass effect, edema, and enhancement. However, a growing astrocytoma of the basal ganglia, pons, or cerebellum, when small, may be indistinguishable from one of these lesions on noncontrast MR—especially in cases when there are numerous, confluent T2 hyperintense foci in expected locations. If a lesion has any suspicious characteristics, it is recommended that the patients have a follow-up contrast-enhanced MR scan, including diffusion-weighted imaging (DWI), perfusion imaging, or proton MR spectroscopy, 6 months to a year after the first study. A tumor will typically show progressive enlargement and mass effect on adjacent structures and often enhances (
Fig. 6-6); it may also have locally reduced diffusion in its solid portion (
95), have increased blood volume (
101), and show elevated choline with decreased creatine and NAA on proton spectroscopy (
69,
102,
103,
104,
105).
T1 hyperintensity is observed in the globi pallidi of patients with NF1, with a different radiologic appearance than the characteristic T2 hyperintense foci: the hyperintensity on T1-weighted images develops after the T2 prolongation appears and persists after T2 prolongation disappears, a time course that suggests that the T1 shortening represents delayed or reactive formation of myelin (
Fig. 6-14) (
86,
90,
98,
106). The distribution of the T1 bright signal can be diffuse within the T2
abnormality or be located along its periphery/rim. (Calculation of T1 values shows that the frontal white matter, caudate nuclei, putamina, and thalami also have some T1 shortening (
86), but it is not apparent to visual inspection in these other areas.) In some patients, the T1 shortening is observed at the time of a first MRI, prior to exposure to gadolinium-based contract agents. However, many of the NF patients with appreciable T1 bright signal changes in the globus pallidus also have bright T1 signal in their dentate nuclei and have undergone multiple contrast-enhanced MRI exams; the T1 bright signal changes in some of these cases, therefore, may be a reflection of gadolinium deposition in the brain (
107), rather than a reflection of an intrinsic abnormality of the myelin.
The overwhelming majority of these characteristic T2 bright lesions resolve over time. However, as with all small, nonenhancing foci of T2 prolongation in the brain,
the biological behavior of white matter lesions in NF1 cannot be predicted with complete accuracy by imaging alone. Benign-appearing lesions can degenerate into neoplasms (
108), and neoplastic-looking lesions can shrink or disappear (
63,
64,
65). Also remember that
focal lesions developing in the cerebral cortex or in the subcortical or deep white matter do not represent characteristic T2 bright lesions of pediatric NF1 (
Fig. 6-8) (
93); if present, these should be followed by sequential imaging, as they may potentially represent lowgrade neoplasms.
Occasionally, the hippocampi of pediatric patients with NF1 can show diffusely increased signal on T2-weighted images, often with increased volume (
Fig. 6-15); this finding can be unilateral or bilateral and is not progressive; it persists throughout childhood and adolescence (
93), in contrast to the characteristic T2 lesions described above that regress during later childhood and adolescence. The nature of the hippocampal abnormality is not known. NF1-related hippocampal T2 changes should not be confused with hippocampal sclerosis (in which case the hippocampal volume is decreased) or with neoplastic infiltration.
Vascular Dysplasia
Dysplasia of the cerebral vasculature is increasingly recognized in children with NF1; it develops at a young age, with prevalence of up to 6% (
109,
110,
111). The incidence in the symptomatic NF population appears to be higher, with a predicted range of 7% to 15% (
112). The known significant incidence of vascular abnormalities in the cerebral vasculature and in other parts of the body makes vascular dysplasia a distinct clinical entity. Any child with NF1 who has seizures, mental retardation, paralysis, or severe headaches may have a vascular dysplasia, particularly in the absence of a brain tumor or hydrocephalus. However, clinically, a majority of NF1 patients with vascular dysplasia do not have focal deficits as a result of their arteriopathy by the time the dysplasia is diagnosed with MRI. Intracranial vascular dysplasias can result as a complication of irradiation for optic nerve or optic chiasm gliomas (presumably radiation arteritis in these cases; see discussion in
Chapter 3).
Common vascular abnormalities include stenoses, occlusions, ectasia, moyamoya disease, and fusiform aneurysm formation. Most commonly, the dysplasia results from intimal and smooth muscle proliferation, with resultant stenosis or occlusion; they can progress overtime (
Fig. 6-16) and may require revascularization surgery (
109,
111). The carotid (common or internal) artery, proximal middle cerebral
artery, or proximal anterior cerebral artery is the most common site of involvement (
Figs. 6-16 and
6-17). Moyamoya phenomenon with marked enlargement of the lenticulostriate arteries (which function as collateral vessels) is seen in many such patients (
113). (Moyamoya is discussed further in
Chapter 12.) Fusiform arterial dilation, cerebral aneurysms, and arteriovenous fistulas are also described, although less commonly (
109,
113,
114).
Vascular dysplasias can be difficult to detect on standard CT or MR images; with MR, close inspection of the arterial signal voids (
Figs. 6-16 and
6-17), especially the distal carotid arteries and the circle of Willis (to detect narrowing of the cavernous or supraclinoid carotids and proximal middle and anterior cerebral arteries), is essential in patients with NF1 to avoid delays in clinical detection of these lesions; note that the
intracranial vessels are poorly seen on FLAIR images, so acquisition and inspection of T2-weighted spin-echo images is strongly encouraged when imaging children with NF1, particularly those with declining function. As many vascular lesions are subtle and easily overlooked, some authors promote the routine use of MR angiography to better delineate intracranial vasculopathy (
17). If vascular dysplasia is suspected, MR (or CT) angiography should be performed; conventional, catheter angiography is usually reserved for a presurgical workup. Angiography generally shows either severe stenosis or occlusion of the involved cerebral vessels (
Fig. 6-17); arterial dysplastic lesions are also demonstrated.
Calvarial and Orbital Abnormalities
Other cranial/intracranial manifestations of NF1 include deficiencies of the sphenoid wing and of the occipital bone along the lambdoid suture.
Sphenoid bone dysplasia allows herniation of the temporal lobes into the orbit (
Fig. 6-18). The pulsations of the temporal lobe may be transmitted through the globe, thus being visible externally as pulsatile exophthalmos (or enophthalmos secondary to atrophy of orbital contents). The globe may be dysplastic, enlarged (buphthalmos), or hypoplastic. Sphenoid wing deficiency is nearly always associated with PNs in the orbit or periorbital regions (
Fig. 6-18). Recent work indicates that the abnormalities of the bony orbit may progress over time, suggesting that these are not simple dysplasias of bone, as previously postulated, but may in part result from erosions by the adjacent neurofibromas (
Fig. 6-18) or optic nerve tumors (
118,
119). Occipital bone dysplasia is usually not clinically significant; it is easily recognized with CT imaging. With MR imaging, it presents as distortion and focal outward bulging of the cerebellar hemisphere deep to the bone defect; this can be confusing until the bone defect is recognized.
Neurofibromas and Plexiform Neurofibromas
Another cause of intracranial complications in NF1 is excessive growth of craniofacial PNs. They tend to progress along the nerve of origin (usually small, unidentified nerves) into the intracranial space, causing distortion and compression of the brain. They most commonly arise from branches of the trigeminal nerve, with occasional intracranial involvement to the level of the cavernous sinus. In the orbit, they act as masses, causing impaired ocular movements and exophthalmos (
Fig. 6-19). Orbital neurofibromas commonly arise in the region of the orbital apex or superior orbital fissure (in the distribution of the first division of the trigeminal nerve); careful evaluation of the images
often reveals extension into the cavernous sinus, nasopharynx, or pterygomaxillary fissure. The neck is another common location for neurofibromas, especially along the course of the vagal nerve, with an estimated occurrence of 25% to 30% in patients with NF1 (
Fig. 6-20) (
120). Neck masses, their appearance, and their differential diagnoses are discussed in
Chapter 7.
Solitary neurofibromas have slightly greater signal intensity than does skeletal muscle on T1-weighted sequences. On T2-weighted sequences, the lesions have variable signal intensity. Most commonly, the periphery of the lesions tends to be of high T2 signal intensity with respect to muscle, whereas the center of the lesions is often of low signal intensity (
121,
122,
123,
124); this has been referred to as the “target sign” (
Figs. 6-20 and
6-21) (
125). The central area of decreased T2 intensity is probably related to the known dense central core of collagen within these lesions (
122,
123). Collagen has a low mobile proton density and therefore is of low signal intensity on T2-weighted images. Enhancement after administration of paramagnetic contrast is variable, although at least a portion of the tumor usually enhances (usually the central aspect, corresponding to the central T2 hypointense (“target”) components (see
Fig. 6-24).
PNs tend to be of low attenuation on CT and generally show little enhancement after administration of intravenous contrast.