Childs Nerv Syst (2015) 31:1661–1676 DOI 10.1007/s00381-015-2834-z

SPECIAL ANNUAL ISSUE

Posterior fossa tumors in children: developmental anatomy and diagnostic imaging Charles Raybaud 1 & Vijay Ramaswamy 2 & Michael D. Taylor 3 & Suzanne Laughlin 1

Received: 2 July 2015 / Accepted: 10 July 2015 # Springer-Verlag Berlin Heidelberg 2015

Abstract Introduction Modern understanding of the relation between the mutated cancer stem cell and its site of origin and of its interaction with the tissue environment is enhancing the importance of developmental anatomy in the diagnostic assessment of posterior fossa tumors in children. The aim of this review is to show how MR imaging can improve on the exact identification of the tumors in the brainstem and in the vicinity of the fourth ventricle in children, using both structural imaging data and a precise topographical assessment guided by the developmental anatomy. Results The development of the hindbrain results from complex processes of brainstem segmentation, ventro-dorsal patterning, multiple germinative zones, and diverse migration pathways of the neural progenitors. Depending on their origin in the brainstem, gliomas may be infiltrative or not, as well as overwhelmingly malignant (pons), or mostly benign (cervicomedullary, medullo-pontine tegmental, gliomas of the cerebellar peduncles). In the vicinity of the fourth ventricles, the prognosis of the medulloblastomas (MB) correlates the molecular subtyping as well as the site of origin: WNT MB develop from the Wnt-expressing lower rhombic lip and have a good prognosis; SHH MB develop from the Shhmodulated cerebellar cortex with an intermediate prognosis

* Charles Raybaud [email protected] 1

Pediatric Neuroradiology, Hospital for Sick Children, 555 University Avenue, Toronto, ON M5G1X8, Canada

2

Neurooncology, Hospital for Sick Children, 555 University Avenue, Toronto, ON M5G1X8, Canada

3

Neurosurgery, Hospital for Sick Children, 555 University Avenue, Toronto, ON M5G1X8, Canada

(dependent on age); recurrences are local mostly. The poor prognosis group 3 MB is radiologically heterogeneous: some tumors present classic features but are juxtaventricular (rather than intraventricular); others have highly malignant features with a small principal tumor and an early dissemination. Group 4 MB has classic features, but characteristically usually does not enhance; dissemination is common. Although there is as yet no clear molecular subgrouping of the ependymomas, their sites of origin and their development can be clearly categorized, as most develop in an exophytic way from the ventricular surface of the medulla in clearly specific locations: the obex region with expansion in the cistern magna, or the lateral recess region with expansion in the CPA and prepontine cisterns (cerebellar ependymomas, and still more intra-brainstem ependymomas are rare). Finally, almost all cerebellar gliomas are pilocytic astrocytomas. Conclusions A developmental and anatomic approach to the posterior fossa tumors in children (together with diffusion imaging data) provides a reliable pre-surgical identification of the tumor and of its aggressiveness. Keywords Posterior fossa tumors . MR imaging . Brainstem glioma . DIPG . Medulloblastoma . Ependymoma Historically, the diagnosis of tumors of the CNS in children is based on a combination of pattern recognition (structural analysis of the lesion) because pathology was, and still is, the gold standard in the definition of the tumors and of topographic assessment because it has been recognized for a long time that in children at least, there is a high degree of correlation between the location of the tumor and its nature. The developments of MR technology (multiple sequences, high definition, perfusion imaging, diffusion imaging/quantitative DTI, magnetic susceptibility imaging) have for years provided

1662

extremely efficient tools to evaluate and characterize the structure of the tumor. More recently, a better understanding of the interaction between the mutated stem cell and its environment is enhancing the significance of anatomy in the specification of the tumor [10, 13]. This is clearly established for medulloblastoma: the molecular classification correlates better than histology both with location and prognosis [27, 31]. In the brainstem, gliomas of the pons are consistently highly malignant, while tegmental gliomas are typically low grade [8]. The aim of this review is to show how MR imaging can improve on the exact identification of the tumors in the brainstem and in the vicinity of the fourth ventricle in children, using a precise topographical assessment guided by the developmental anatomy in addition to structural imaging data. It addresses the Bclassic^ posterior fossa tumors only, not the more unusual ones such as PNET/ATRT or hemangioblastomas.

Structural imaging Conventional imaging includes T1, T2, and FLAIR sequences. Based on the fact that T1 and T2 relaxation times depend on the molecular environment of protons, they discriminate tissues with different chemical composition: white and gray matter, different gray nuclei, etc. Tumors obviously have a different structure from normal tissue: no or less myelin, more water, uneven cellular density, necrosis, and/or hemorrhages. Different compartments of water may be differentiated: total water (T1, T2) and water bound to macromolecules only (FLAIR). In addition, diffusion imaging explores the extracellular water. Because T1, T2, and to some extent FLAIR sequences demonstrate clearly the contrast between different tissues and between parenchyma and CSF, they provide superb anatomical images. Modern equipment allows high definition anatomic imaging: 3 mm for the classic T2/ FLAIR sequences, 1 mm for 3D-T1 acquisitions, and 0.5 mm for steady-state T2 images (with loss of intra-tissue discrimination capability however). Diffusion-weighted imaging/average diffusion coefficient (DWI-ADC) evaluates the degree of freedom of motion of the water molecules in the extracellular spaces (diffusivityADC). Maximum in CSF, diffusivity in tissues reflects the cellular density, the intra- or extracellular edema, etc. It is presently the most significant structural parameter when assessing brain tumors: high in low-grade glioma (dark on DWI, bright on ADC images), low in tumors with high cellularity, and high nuclear-cytoplasmic ratio (bright on DWI, dark on ADC images). Therefore, it allows a clear distinction between glioma on one hand and medulloblastoma or ependymoma on the other hand, but not really between medulloblastoma and ependymoma due to a significant overlap of the diffusivity metrics [16, 38]. Diffusion tensor imaging (DTI) provides a tri-dimensional quantification of the

Childs Nerv Syst (2015) 31:1661–1676

diffusivity, and because spatial diffusion depends on the morphology of the space in which it is evaluated, it reflects the structural directionality (fractional anisotropy FA): this provides tractography images. Contrast enhancement Normal brain tissue is poorly vascularized (blood volume is only 4 % of brain tissue) and the normal blood–brain-barrier BBB prevents nonphysiologic molecules from entering it. Contrast enhancement may reflect either an enlarged vascular bed (e.g., hemangioblastoma) or a leaky BBB (e.g., pilocytic astrocytoma), or both (e.g., glioblastoma multiforme). Perfusion imaging evaluates the blood perfusion within the tumor itself (high in malignant tumors, low in non-vascular benign ones) [20], but this is not available on every clinical scanner. Evaluation of a BBB leak is still less readily available. Magnetic susceptibility imaging (T2*GE, SWI etc.) reflects local magnetic field changes induced by iron-containing products: tiny foci of hemorrhage often associated with malignancy, hemosiderin in macrophages [19, 41, 42], as well as calcifications. Specific sequences may also nicely demonstrate deoxyhemoglobin-containing veins [36]. Proton MR spectroscopy (1HMRS, MRS) uses the influence of the molecular environment on the resonant frequency of the protons to differentiate specific molecular compounds along a frequency spectrum (location evaluated in ppm) and evaluate their relative amount from the relative size/area of the peaks: NAA (neuroaxonal activity mostly) at 2.01 ppm, creatine (energy metabolism) at 3.02 ppm, choline (cellular turn-over) at 3.22 ppm, myo-inositol (a marker of the glia) at 3.56 ppm, as well as lactate (energy failure, necrosis) at 1.33 ppm, lipids/ macromolecules (necrosis) at 0.9–1.3 ppm, or taurine (a marker of embryonal tumors) at 3.35 ppm [5, 17, 28, 29]. MRS may be performed as single volume (voxel), or may be multivoxel (Chemical Shift Imaging—CSI) [18]. Very sensitive to magnetic susceptibility artifacts, MRS may not always be optimal in clinical conditions, when tumors are close to the brain/bone interface, in hemorrhagic tumors or in patients with dental hardware or braces.

Anatomic imaging From a practical MR imaging point of view, a few anatomical landmarks are important to recognize when assessing the anatomy of the hindbrain. The midline sagittal cut shows the tegmental cap of the midbrain forming the 3rd ventricular floor behind the mamillary bodies and the smooth sweep of the Sylvian aqueduct and the quadrigeminal plate, the deep indentation of the interpeduncular fossa, and the clearly apparent inferiorly oblique boundary between the midbrain and the pons. More caudally, the wedge-shaped pons is prominent, with a rounded ventral part (with pontine nuclei) well

Childs Nerv Syst (2015) 31:1661–1676

demarcated from the posterior tegmentum pontis (with cranial nerve nuclei and reticular formation); it is limited caudally by the indentation of the pontomedullary sulcus and the superiorly oblique pontomedullary boundary. The Bopen^ cranial portion of the medulla corresponds to the inferior half of the fourth ventricle, with the obex covering its caudal tip. The Bclosed^ caudal medulla looks more like the cord. Behind the pons and medulla, the fourth ventricle forms a triangle between the aqueduct rostrally, the foramen of Magendie caudally, and the fastigial angle indenting the overlying dorsal vermis. The choroid plexus is transversely attached to the inferior vermis, with the CSF space of the vallecula and cistern magna limited by the inferior vermis, dorsal medulla, and supra-occiput. More lateral cuts show the three cerebellar peduncles optimally. Coronal cuts show the whole extent of the brainstem from the thalami to the cord, the three cerebellar peduncles, and the cerebellar hemispheres and vermis. The axial cuts span the whole neuraxis from the upper midbrain to the upper cord. The upper cuts demonstrate the aqueduct and periaqueductal gray matter, the inferior and superior colliculi, the red nuclei and substantia nigra, and the diverging cerebral peduncles. The cut through the rostral pons crosses the oblique midbrain-pons interface, showing the pons anteriorly and the midbrain tegmentum posteriorly, as well as the superior cerebellar peduncles. The more caudal cut shows the ventral pons which contains the pontine nuclei with longitudinal and transverse axonal bundles, tegmentum pontis which contains the nuclei of the pontine cranial nerve and reticular formation, brachium pontis (or middle cerebellar peduncle) on either side, and the ventricular floor with its median sulcus and medial eminences (colliculi faciales). The vermis posteriorly encroaches upon the ventricle, with the hemispheres extending laterally. The cut through the pontomedullary sulcus shows the lateral recesses, inferior cerebellar peduncles, flocculi, and CP angle cistern. The cut below through the Bopen^ rostral medulla shows the anterolateral prominence of the inferior olivary nuclei, the medullary tegmentum, the caudal fourth ventricle with the inferior cerebellar peduncles on either side, and the inferior vermis and cerebellar hemispheres behind. The last cut through the Bclosed^ caudal medulla shows the anterior protrusion of the pyramidal decussation, a tiny ependymal canal, and dorsally the vallecula and cisterna magna.

Development of the brainstem Craniocaudal segmentation (Fig. 1a) Anatomically, the brainstem comprises three segments, the midbrain, pons, and medulla. Yet embryologically, the rostrocaudal segmentation is much more complex. In early development, the hindbrain is subdivided in seven rhombomeres r1-r7 by the expression of

1663

various Homeobox genes. These rhombomeres are clearly individualized: because of specific cell affinities or adhesion properties, cells born within one rhombomere cannot cross the intersegmental boundaries to mix with cells of other rhombomeres [50]. Clearly defined by the expression of the Fgf8 molecule, the mesencephalo-rhombencephalic isthmus itself is identified as a specific segment r0; it is called the isthmic organizer because it influences the regional differentiation [25]. The pons (with most of the cerebellum) corresponds to r1, while the rostral Bopen^ medulla corresponds to the succession of three pairs of segments r2-r3, r4-r5, and r6-r7, each corresponding to the motor innervation of a branchial arch by CN V, VII, and IX, respectively, [56]. The caudal Bclosed^ medulla comprises the Bpseudorhombomeres^ r8-11, corresponding to somites 1-5 [3] (four occipital and one proatlantal), with the rhombo-spinal boundary at the center of somite 5 [3], that is the occipito-atlantal space and C1 spinal nerve (pro-atlantal cleft). Although intersegmental boundaries become permissive to the migration of neural progenitors and axons later in development, this basic brainstem segmentation may still be significant to understand the anatomic specificities of brainstem gliomas. Ventro-dorsal patterning Ventrodorsally, the neural tube is made of a basal plate (motor), an alar plate (sensory) and, in its cephalic segments, a dorsal roof plate (integrative). In the rhombencephalon, the diamond-shaped fourth ventricle is lined by the dorsal edge of the alar plates forming the upper rhombic lips (URL) at the pontine level, and the lower rhombic lips (LRL) at the medullary level (Fig. 1a). A roof plate, the cerebellum, develops at the pontine level; there is no medullary roof plate, but a tela choroidea with the ventricular outlets. The early cerebellar plate (anterior vermis) develops from the roof and alar plates of the isthmus (r0) and bridges the rostral angle of the diamond so that the URL becomes progressively transverse, forming the caudal margin of the cerebellar plate (Fig. 1b) [24, 43, 55]. The posterior vermis (roof plate r1) develops medially from this rostral cerebellar plate, while the cerebellar hemispheres develop laterally from the URL (alar plates r1) (Fig. 1c) [24]. The LRL (with the tela choroidea and ventricular outlets) correspond to the open medulla and are segmented accordingly (alar plates r2-7). There is no roof plate or tela choroidea at the caudal medullary level (r 8-11). Histogenesis of the hindbrain (Fig. 1d) The production of neural progenitors to the brainstem and cerebellum is complex. Neuroblasts from the ventricular zones of the basal plates (motor) and of the alar plates (sensory) remain confined within their own rhombomeres and form the cranial nerve nuclei in the tegmental (dorsal) part of the brainstem, together with part of the reticular formation [47]. Laterdeveloping gray matter (tectal plate, cerebellum, pre-

1664 Fig. 1 Development of the hindbrain. a The diamond-shaped rhombencephalic roof is divided by the transverse choroid plexus in two halves. The cranial half corresponds to rhombomere 1, and the caudal half corresponds to rhombomeres r2–7. The edges of the cranial half are the upper rhombic lips (UPL) and the edges of the lower half are the lower rhomcic lips (LRL). Pseudorhombomeres 8–11 form the caudal medulla. b The early cerebellar plate (upper vermis) develops from the roof and alar plates of the isthmic organizer r0 (open arrows). c The lower vermis develops from the cerebellar plate (white arrow), and the cerebellar hemispheres develop from the URL (blue arrows). The neuroblasts of the ventral pontine nuclei (PN) and inferior olivary nuclei (ION) migrate from the LRL. d Sagittal cut to illustrate the origin of the main gray matter features of the hindbrain: cranial nerve nuclei (CN) from the tegmental subventricular zone (SVZ); cells of Purkinje and dentate cells from the cerebellar SVZ; external granular layer (EGL) from the URL; and PN and ION from the LRL

Childs Nerv Syst (2015) 31:1661–1676

a

b

Cbl plate

MB r0 URL

r0

r1

URL

Ch Pl

r1

Ch Pl

r2-3

r2-3

r4-5 LRL

r4-5 LRL

r6-7

r6-7

r8-11

c Cbl plate

r8-11

d r0

PC

EGL

DN

URL to CN

r1 to PN r2-3

ION (from LRL)

r4-5 to ION

PN (from LRL)

r6-7 r8-11

cerebellar nuclei) is composed of neuroblasts which migrate regardless of the inter-neuromeric boundaries. Neuroblasts from the roof plate and alar plate of the isthmic organizer r0 migrate rostrally to form the tectal plate and caudally to form the early cerebellar plate (upper vermis), respectively. A germinative ventricular zone develops on the ventral aspect of the

cerebellar plate and produces the GABAergic neural progenitors of the deep cerebellar nuclei, the Purkinje cells and the interneurons, as well as the Bergman glia [24, 55]. The URL is another germinative zone which produces glutamatergic neuronal progenitors which migrate to the surface of the cerebellum and form a secondary germinal zone there, the external

Table 1 Brainstem tumors (excluding midbrain tumors)

N

Neuro-pathology

HGG

LGG

PNET

Pons Brachium pontis Medullary and ponto-medullary, tegmental Medullary, lateral

63 12 17 6

10/63 12/12 14/17 4/6

55 1 1

5 11 16 6

3

Medullary, anterior Cervico-medullary

2 8

1/2 6/8

2

Brainstem, diffuse Total

6 114

2/6 49/114

3 62

(2 GGG)

Bold numbers emphasize the significant groups

8 (2 GGG) 3 49

3

Childs Nerv Syst (2015) 31:1661–1676

granular layer (EGL); until the second postnatal year, the EGL produces granules cells which migrate to the internal granular layer [55]. The SHH signaling pathway modulates the production of granule cells, and SHH expression characterizes a specific subtype of medulloblastoma which develops from the surface of the cerebellum [9, 39]. More caudally, the LRL is still another germinative zone which produces neuronal progenitors to the precerebellar nuclei: inferior olive, basal pontine nuclei and reticular formation as well as much later, the cochlear nuclei from the anterolateral region of the LRL [55]. This region has a strong WNT expression, and WNT expression characterizes another specific subtype of medulloblastoma which develops in the lateral recesses of the fourth ventricle [9, 39].

Hindbrain connectivity Gray matter is connected, and because of the complexity of its organization, the brainstem is a complex major crossroad of axonal bundles. It has been assumed that the pattern of extension of the glial tumors within the brainstem would be dependent on the fascicular anatomy [8]. Beside the Bfacilitating^ descending and ascending white mater tracts, transverse decussation planes would create Bbarriers^ preventing the extension of benign tumors: pyramidal decussation, internal arcuate fibers, and olivocerebellar tracts would divide the medulla; pontocerebellar fibers and trapezoid body would isolate the pons from the rostral Fig. 2 Ventral pons, DIPG. Extensive, heterogeneous mass centered in the pons with anterior exophytic components surrounding the basilar artery (a–b), with no diffusion restriction (c) and with prominent pontine veins (d)

1665

medulla; and pontocerebellar fibers and decussation of the superior cerebellar peduncle would isolate the midbrain from the pons [8]. Benign tumors, but not malignant ones, would respect those barriers [8]. Modern developmental perspectives rather suggest the role of cellular and rhombomeric developmental specificities to explain the tumoral specificities.

Diagnostic application Brainstem tumors (excluding midbrain tumors) Pontine tumors The pons is the most common location of the brainstem tumors, accounting for 63/114 (55 %) in our experience (Table 1). Pontine tumors were almost always malignant: 55 were highly malignant DIPG (diffuse infiltrative pontine glioma), and three were PNET. Pathology was obtained in 10 cases only, and diagnosis for the remainder was based on location, appearances and evolution. Within the pons, the location characteristically was ventral (Fig. 2a–d), or ventrolateral (Fig. 3) but not as far lateral as the brachium pontis. Initially at least, tumors seem to expand the pons craniocaudally and compress the adjacent midbrain and medulla, before invading them. Most DIPG were exophytic anteriorly or anterolaterally in the prepontine cistern and surrounded the basilar artery without really invading it. Yet,

1666

Childs Nerv Syst (2015) 31:1661–1676

course. In only 5/63 cases in our series (all laterally located), a pontine tumor was found to be (and to remain) benign based on histology (3/5) or a very long survival. Finally, three cases of PNET were diagnosed histologically (2/3) or because of Bspecific^ MR features.

Fig. 3 Lateral pons, DIPG

3/55 were exophytic dorsally (Fig. 4a–b) and 3 remained central, expanding the pons concentrically without any exophytic component. Hydrocephalus was unusual. The characteristic appearance of the DIPG is that of a malignant glioma. Essentially, all present with a diffusely low signal on T1 and bright signal on T2 and FLAIR sequences compared with the normal parenchyma. Malignancy is expressed by a few particular features: in a diffusely high T2, high ADC lesion, there are foci of lower T2 intensity and of mildly (relative to the surrounding tumor) restricted diffusion, reflecting a high cellular density [4], poorly defined limits (better appreciated on FLAIR), surrounding edema, invasion of adjacent segments, hemorrhages (especially multiple tiny diffusely distributed petechial hemorrhages) [2, 9, 52], and necrosis. Enhancement is extremely variable, from absent to faint, patchy, or ring-like surrounding a necrotic areas; special MR techniques may be needed to demonstrate it [4]. Perfusion would be high [20]. Even in non-enhancing tumors, the prominence of the intrinsic pontine vessels reflects a high blood flow, that is, malignancy (Fig. 2d). Obviously, leptomeningeal dissemination is evidence for high grade. In one case, the tumor in the lateral pons became malignant after several years of an apparently benign Fig. 4 Posterior pons, DIPG

Tumors in the brachium pontis A tumor apparently centered in the brachium pontis (rather than the lateral pons, which may be a subjective appreciation given the associated anatomic distortion) was found in 12 cases in our series. Of these, 11 were identified by neuropathology as low-grade gliomas (LGG), and one as a high-grade glioma (HGG). LGG are typically well demarcated from the surrounding parenchyma, they cross anatomic boundaries only rarely, and present as solid masses without or with a cystic component (Fig. 5). They commonly bulge dorsomedially into the fourth ventricle but also anterolaterally in the cisterns. Their signal is low on T1, high on T2 and FLAIR, and no restricted diffusion is shown. The solid component tends to enhance more diffusely and homogeneously than in HGG. On follow-up, the growth of the mass is slow, hardly apparent on short intervals, reflecting a usually long clinical history. In the only case of HGG (documented by pathology), the tumor was centered far laterally, protruded into the cerebellopontine angle cistern, and presented heterogeneous low T1, high T2 and high FLAIR signal, faint diffusion restriction, and incomplete enhancement with necrotic features; survival time was only four months. Tegmental brainstem tumors Tumors originating from, and remaining in the dorsal aspect of the brainstem (tegmentum), were observed in 17 of our 114 brainstem tumors. They may develop from the low medulla at the medullocervical junction caudal to the fourth ventricle (three cases, perhaps minor forms of cervicomedullary tumors), or from the medullary part of the floor of the ventricle (five cases), or more commonly from the medullary and most or all of the pontine part of the floor of the ventricle (nine cases) (Fig. 6a–b). Except for the two small low medullary tumors, pathology was obtained

Childs Nerv Syst (2015) 31:1661–1676

1667

All were of small to moderate size, growing slowly, with low T1, high T2 and high FLAIR signal, no restricted diffusion, and enhancement of the lateral exophytic component.

Fig. 5 Brachium pontis, LGG

from surgical resection tissue and disclosed a LGG in 14/15 cases, and a HGG in one 22-month-old girl. Hydrocephalus is associated with large tumors (11/17). On the first diagnostic study, the tumor is structurally almost always solid (16/17), with low T1, high T2, and FLAIR, and no restricted diffusion; one was both solid and cystic. It is typically rather homogeneous, but when it is very large, its center may appear necrotic-looking. Enhancement is variable: partial, faint, and superficial in small masses, more prominent in large tumors, sometimes with a necrotic-looking center. The degree of infiltration of the brainstem by LGG is variable also, but remains partial and dorsal. The structural features of the HGG were different as the mass appeared poorly defined, roughly iso-intense on T1, T2, and FLAIR, with intermediate ADC, and a very faint enhancement only; the brainstem infiltration was extensive. Lateral medullary tumors In 6 of our 114 patients with brainstem tumors, the tumor originated from the inferior cerebellar peduncle in the rostral medulla and expanded laterally into the CPA cistern (Fig. 7a, b). Pathology was available in 4/6: all four were diagnosed as LGG, two specifically as gangliogliomas (GGG); only one (a GGG) was partly cystic. Fig. 6 Tegmental LGG. A large dorsal mass arises from the medullary and pontine tegmentum and extends into the fourth ventricular lumen, with little brainstem infiltration (a sagittal T1; b axial FLAIR)

Anterior medullary tumors Tumors in the anterior medulla are rare: 2/114 cases in our series. In one of our cases biopsy initially disclosed a grade II astrocytoma (Fig. 8a), but the tumor kept growing in spite of the chemotherapy, and 4 years after diagnosis, metastases developed in the spinal canal and in the lateral ventricles (Fig. 8b). The other developed from the rostral medulla and pontomedullary sulcus, with a deep implantation in the medulla and a significant anterior exophytic component encasing the terminal portions of both vertebral and the proximal basilar arteries. There was no pathology but supratentorial ependymal metastases developed very quickly, and survival was less than 13 months, so that it was assumed to be a HGG. Cervico-medullary tumors The specificity of the cervicomedullary junction tumors has been recognized for a long time [7]. In the present series, they represented 8/114 cases, with a variable extension caudally and cranially. The smallest one extended from the obex down to C2. Largest extensions were cranially to the midpons, and caudally to the C6 level. In all, the upper and lower limits were clearly demarcated. In two cases, the mass was mildly lateralized and invaded one inferior cerebellar peduncle, and for one, the white matter of the cerebellar hemisphere as well. All presented a smooth, but typically voluminous posteriorly exophytic component into the cisterna magna (Fig. 9a, b). All infiltrated the whole thickness of the upper cord, and part of, or more commonly, the whole thickness of the medullary segment involved (Fig. 9a, b). All tumors but one were fully solid; one was solid anteriorly with a prominent anterior exophytic component and a posterior cystic component which was filling the cisterna magna. Pathology obtained in 6/8 cases demonstrated 4 LGG and 2 GGG. (Of the two remaining cases, one is lost for follow-up (consult), the other one was followed and has been stable from 2007 to date [2015]). Structural appearances

1668

Childs Nerv Syst (2015) 31:1661–1676

Fig. 7 Lateral medulla, LGG.A laterally exophytic mass originates from the left inferior cerebellar peduncle (a axial FLAIR; b coronal T2)

were in keeping with a low-grade tumor in all cases: hypo T2, hyper T2 and FLAIR signals, no diffusion restriction; all enhanced, variably. The only anatomic plane apparently respected by the medullary extension of the tumor was the ponto-medullary junction, in 4/8 cases, but in one case, the tumor extended in the pons. As mentioned above, one inferior cerebellar peduncle was invaded in two cases. Diffuse tumors of the brainstem In 6 of our 114 cases, the tumor involved the brainstem, diffusely. From the structural appearances, duration of follow up and in two cases, pathology, the tumor was benign in three cases (Fig. 10a, b), and malignant in three. It infiltrated the whole thickness of the brainstem in all. In one HGG case, the tumor appeared bisegmented, one mass in the pons and one in the medulla (Fig. 11a, b), in contrast with other diffuse tumors in which the ponto-medullary sulcus was effaced. In one patient with a previous history of ependymoma, the histologically proven LGG is assumed to have been radiation induced. Midbrain tumors As compared with the tumors of the pons and medulla, tumors centered in the midbrain are uncommon: 19 cases only, if the mostly pontine or mostly thalamic tumors are excluded. Such Fig. 8 Anterior medulla, HGG. Initial imaging shows a well circumscribed anteriorly exophytic mass (a sagittal T2); histology disclosed a LGG grade II. The tumor kept expanding over the years and became massively metastatic (b sagittal T1 postcontrast)

a small number precludes any generalization, and examples of all grades I–IV have been reported [6, 37, 44, 52]. The vast majority however seems to be LGG: only two patients had a malignant lesion, one peduncular HGG and one tectal PNET. Topographically, the tumors may be classified as tectal (7 cases, 6 LGG and one PNET), tegmental (6 cases, all LGG), or peduncular (six cases, five LGG, and one HGG). Tectal plate tumors are sometimes difficult to differentiate from dormant Bhamartomas^ of the periaqueductal region [1, 23, 40] or of the quadrigeminal plate [1, 14, 48], and the common tegmento-tectal or tegmento-peduncular diffusion makes a ventro-dorsal classification uncertain also.

Medulloblastoma Medulloblastoma (MB) is an embryonal neural tumor, histologically classified in five groups by the WHO [21]: classic MB, anaplastic MB, large cell MB, desmoplastic-nodular MB, and MB with extensive nodularity. Practically, a simpler classification in three groups is used: classic MB (the most common), desmoplastic-nodular MB, and large cellanaplastic (LCA) MB, the most malignant. This classification correlates relatively well, if by no means rigorously, with structural MR imaging [57]. Classic MB presents as a central

Childs Nerv Syst (2015) 31:1661–1676

1669

Fig. 9 Cervico-medullary, LGG. Extensive tumor with posterior exophytic component; it infiltrates the full thickness of the cord (a, b) and of the involved segment of the medulla (b, T1 post-contrast)

posterior fossa mass filling the fourth ventricle (and associated with high-pressure hydrocephalus), fairly compact and homogeneous, well circumscribed. Signal is moderately low on T1, only mildly bright, if at all, on T2 (would be mildly hyperattenuating on CT), moderately bright on FLAIR, and highly restricted on DWI/ADC. This reflects the high cellularity of the tumor and its high nuclear to cytoplasmic ratio. Focal hemorrhage or cystic necrosis may be seen. Enhancement is variable: most tumors do enhance, in a diffuse, patchy or ring-like manner, but some do not in spite of the fact that they still may present with prominent vascular signal voids. Macroscopic intracranial and/or spinal leptomeningeal dissemination may be demonstrated, but not commonly on the first examination. Desmoplastic-excessive nodularity MB is located peripherally in the cerebellum; perfectly demarcated, it appears compact and highly restricting, and often somewhat multinodular. Above all, it enhances very avidly, in a multinodular manner, after contrast administration. Large cell-anaplastic MB is the most malignant. Like the other subtypes, it presents with a low T1, slightly elevated T2 and FLAIR only, marked restricted diffusion, and usually diffuse enhancement. In its most typical, although not most common presentation, the tumor in the fourth ventricle is small, often without hydrocephalus, but already associated with early

Fig. 10 Diffuse brainstem, LGG. Extensive, poorly demarcated, non-enhancing mass in the pons and medulla (a sagittal T2; b sagittal T1 post-contrast)

leptomeningeal dissemination. However, LCA MB often present with a non-specific MR pattern. In recent years, the perspective on these tumors has changed with the development of the modern genomic approach: distinct molecular variants of MBs are recognized with different evolutive potential [9, 15, 26, 27, 34, 46]. This new approach might offer the possibility to adapt treatments to the aggressiveness of a specific tumor in a specific patient and to attenuate the often severe secondary effects of therapy on the developing brain. Because different molecular expressions may relate to different anatomic origins (see above, Histogenesis of the hindbrain), anatomically oriented MR imaging may correlate well the molecular classification [31]. Currently, four subgroups are identified: WNT subgroup, sonic hedgehog subgroup (SHH), group 3, and group 4 (these last two subgroups lack specific molecular markers, but still, tumors in each group cluster together). WNT MB is characterized by a good long-term prognosis. It presents mostly with the classic histology, affects girls and boys equally, but is uncommon in infants [46]. It often seems to develop from the lower rhombic lip which has a strong Wnt expression [9] and therefore may be identified when it is centered on a lateral recess of the fourth ventricle. Similarly, SHH signaling modulates the development of the external granular layer and the

1670

Childs Nerv Syst (2015) 31:1661–1676

Fig. 11 Diffuse brainstem, HGG. Extensive tumor with two nodules centered in the pons and in the medulla, each with its own appearance (a sagittalT2; b sagittal T1 post-contrast)

production of granule cells; accordingly, tumors of the SHH MB subgroup involve the cerebellar cortex. They may be of any histologic subtype (essentially all desmoplastic/nodular, but also classic and LCA), with a good prognosis in infants, but poor prognosis in older children likely due to cooccurrence of TP53 mutations [35, 58]. Tumors of both group 3 and group 4 develop in the immediate vicinity of the fourth ventricle, may be histologically classic or LCA, with common metastatic dissemination. Group 3 is found in infants mostly, is associated with MYC amplifications and has the worst prognosis. Group 4 corresponds to the archetypal MB, with an intermediate prognosis. Interestingly, its prominent MR feature is the absence of enhancement after contrast administration [31], implying a specific Bpreservation^ of the tumoral BBB. At recurrence, SHH MB recur most frequently locally in the resection cavity, whereas group 3 and group 4 recur almost exclusively with leptomeningeal dissemination [33]. In practice, things are not as simple as could be expected, assumedly because other developmental factors are involved, which are not known. From a series of 126 MB which were classified in subgroups in our institution, the initial preoperative MR could be retrieved in 110 cases (Table 2). The WNT group included 13 tumors; 7/13 were located in the Table 2

region of the lateral recess/CP angle (Fig. 12a, b); 6/13 were mostly intraventricular: a possible explanation is that from the lower rhombic lip, the tumor may develop medially rather than laterally and may therefore be confused with a ventricular tumor. In such a case, as WNT-MB are primarily brainstem tumors, assessing the attachment of the lesion to the dorsal brainstem might contribute to its identification (this was not assessed in our retrospective series). The SHH group included 24 tumors, of which 19 (79 %) were clearly cortical (Fig. 13a–d); hence, 21 % were not identified as such. This may be because they originate in the floccular cortex in the CP angles, or in the nodular cortex bulging in the inferior fourth ventricle, respectively. Group 3, which is made of younger children, has the worst prognosis. It is interesting as it is radiologically clearly heterogeneous: out of 27 cases, 4/27 developed not within but adjacent to the fourth ventricle (Fig. 14a, b): this never occurs in the other groups and may relate to the tumor originating from the GABAergic germinative ventricular zone (in 23/27 it appeared classically located within the ventricle). Also, exclusively in this group, the tumor may be quite small at the time of the first examination (longest diameter less than 35 mm). This was observed in 9/27 patients, typically with

Medulloblastoma

Fourth ventricle: 67 % Para-ventricular Lateral recess Peripheral Small tumor size (longest diameter <35 mm)

Wnt 13/110

SHH 24/110

Group 3 27/110

Group 4 47/110

6

2

39

7 (54 %)

3 19 (79 %)

21 4 (15 %) 2 9 (33 %) (with mets 6/9)

Poor/no enhancement

3

Metastases

4 (17 %)

Bold numbers emphasize the significant groups

5 3

38 (81 %) 7 (26 %) (intracranial 5/7) (with small tumor 6/7)

11 (23 %) (intracranial 6/11)

Childs Nerv Syst (2015) 31:1661–1676

1671

Fig. 12 Medulloblastoma, WNT subgroup. Large mass with low T2 signal (a) and restricted diffusion (b) centered into the left lateral recess

no ventriculomegaly, and was characteristically associated with an early leptomeningeal or ependymal dissemination in 6/9 cases (Fig. 15a, b). This correlated with LCA histology in four cases, but with classic and desmoplastic histology in four and one cases, respectively. Group 4 was the largest group (47/110, or 43 %), with the most typical presentation (large tumor with low ADC centered in the fourth ventricle, associated with hydrocephalus and not uncommonly, early dissemination), but characteristically in 81 % of the cases, there is no or only a very partial enhancement (Fig. 16a–c). As was mentioned above, this suggests an Bintact^ BBB, an unexpected finding for a rather malignant embryonal tumor. Fig. 13 Medulloblastoma, SHH subgroup. Compact, heterogeneous mass developed peripherally in the right hemisphere (a, b), surrounded with edema (b), with restricted diffusion (c), and avid post-contrast enhancement (d)

Posterior fossa ependymoma Ependymoma is the third most common tumor of the CNS in children after the glioma and the medulloblastoma. It occurs in young children mostly (50 % before age 5), develops in the posterior fossa in 2/3 of the cases, typically in the vicinity of the fourth ventricle, and therefore is associated with hydrocephalus. Histologically, it may be a grade II (ependymoma) or a grade III (anaplastic ependymoma) tumor, but it is unclear whether this affects the prognosis [11, 49]. This tumor is solid, and from its attachment to the ventricular lining (usually), it develops in the ventricle and the cisterns where remarkably it adapts to their shape (CPA cistern, cisterna magna, upper

1672

Childs Nerv Syst (2015) 31:1661–1676

Fig. 14 Medulloblastoma, group 3 (juxtaventricular subtype). The centrally located mass (a) is actually juxtaventricular, rather than intraventricular (the wall of the fourth ventricle is intact on b)

cervical arachnoid spaces). It does not really invade the brainstem but rather displaces and compresses it. Recent data show that the stem cell of ependymoma is the radial glia cell, and because radial glia have specificities that depend on where they develops [45], ependymomas have different specificities depending on whether they grow in the cerebral hemispheres, in the posterior fossa, or in the spinal cord [32]. Within the posterior fossa itself, two groups are differentiated on the basis of age, location, and histologic predictors of prognosis, as well as on a molecular basis: laterally located ependymoma (group A) do not fare as well as medially located ones (group B) [45, 53]. Recent genomic advances have revealed that posterior fossa ependymoma comprises two distinct molecular variants termed EPN-PFA and EPN-PFB [22, 30, 54]. EPN-PFA are more common in infants and have a poor prognosis whereas EPN-PFB tend to occur in older children and adults and are associated with an excellent prognosis, partially explaining the prognostic value of age at presentation. Overall in children, it is agreed upon that only the completeness of resection and the age of the patient have a prognostic value. Structural MR imaging is not really helpful, as the characteristics of the lesion are not specific: low T1, high T2 and FLAIR signals; high cellularity resulting in restricted diffusion, possibly more so in the anaplasic subgroup [16, 38], but the range of measured diffusivity values, and the overlap Fig. 15 Medulloblastoma, group 3 (aggressive subtype). The tumor is relatively small, with no hydrocephalus (a). Contrast administration however discloses early leptomeningeal dissemination in the posterior fossa cisterns (b)

with other hypercellular tumors such as the medulloblastoma, limit the usefulness of this approach in individual patients. Enhancement after contrast administration is variable. Hemorrhages, necrosis, and calcification may also be found in other types of tumors. A precise topographic assessment therefore may well provide the most salient diagnostic information. More than two decades ago, a careful correlation between the postoperative survival and the microanatomic localization of the ependymoma in the posterior fossa was proposed, based on surgical data [12]. The authors distinguished three categories depending on the site of origin of the tumor. In the midfloor type, the tumor originates from the inferior part of the fourth ventricle; from there, it expands into the ventricular lumen, across the foramen of Magendie into the cisterna magna and further along the posterior surface of the upper spinal cord. In the lateral type, the tumor originates from the inferior cerebellar peduncle in the lateral recess of the fourth ventricle, and from there, may expand toward the cerebello-pontine angle and into the fourth ventricle. Finally, in the roof-type, the tumor arises (according to the authors) from the inferior medullary velum and grows in the cerebellum. An MR imaging study using the same anatomic criteria and with surgical correlation in 25 patients from our institution was reported previously [51]. It confirmed that infratentorial ependymomas could be classified as

Childs Nerv Syst (2015) 31:1661–1676

1673

Fig. 16 Medulloblastoma, group 4. Huge posterior fossa tumor with hydrocephalus (a) and restricted diffusion (b) but no tumoral enhancement after contrast administration, despite the prominent vessel seen in the center of the mass (c)

lateral type and midfloor type (there was no roof-type tumor in that series), defined radiologic criteria for each type, and concluded that the lateral-type tumors have a significantly increased risk of residual tumor compared to the mid-floor type because of the encasement of the nerves and arteries, and of the difficulty of approaching surgically the components of the tumor that extend in the prepontine cistern. With the perspective of identifying consistent molecular markers for each location, this study was expanded and currently includes 52 patients, with 25 tumors belonging to the midfloor (obex) type, 24 tumors belonging to the lateral recess type, and one tumor belonging to the roof-type. In two remaining cases, it was difficult to decide whether the

Fig. 17 Ependymoma, midfloor (obex) subtype. The mass develops from the caudal tip of the ventricle causing a Bflaring^ of the obex, and extends primarily into the cisterna magna (a). It is strictly median and splays the inferior fourth ventricle open (b)

tumor was at the obex rather than in the roof (one case) or at the obex rather than in the lateral recess (one case). The mean age of the patients with a tumor in the lateral recess is 3 years and 5 months, ranging from 5 months to 8 years and 9 months. The mean age of patients with a tumor at the obex is 6 years and 11 months, ranging from 1 year and 4 months to 17 years. Even though the group with lateral recess ependymoma is clearly younger than the group with obex ependymoma, with infants involved more often, the two groups overlap widely. On MR images, the tumor subtype is recognized above all from its mass effect. When the tumor develops from the obex (midfloor type) the brainstem is displaced ventrally and

1674

Childs Nerv Syst (2015) 31:1661–1676

Fig. 18 Ependymoma, lateral recess subtype. The mass develops in the CP angle cistern, and distorts the brainstem (a); in this case, it remains fully extraventricular (b). It extensively encases the corresponding vertebral and basilar artery segments

therefore remains aligned on the sagittal images (Fig. 17a, b). The tumor is attached to the dorsal brainstem at the obex, which may be flared, or have become indistinguishable because of the tumor. Yet, the dorsal brainstem is not infiltrated, the tumor being strictly exophytic. The mass expands dorsally in the lower fourth ventricle, pushing the vermis, fills the cisterna magna, and often prolapses in the upper cervical canal behind the cord; it may also expand upward to the level of the aqueduct. On axial cuts, the inferior fourth ventricle is splayed and when large enough, the tumor may proceed on either side of the medulla. Vascular encasement is unusual. This growth pattern is not different from the developmental pattern of the tegmental medullary gliomas, which do not always infiltrate the brain. However, differentiation is easy to make because brainstem gliomas always have a high diffusivity (high ADC), whereas ependymomas always have an intermediate or low diffusivity (low ADC). On the contrary, when the tumor develops from a lateral recess (lateral type), it pushes the brainstem to the opposite side, so that the alignment is lost and the tumor appears best on the axial images (Fig. 18a, b). The site of attachment of the tumor itself is usually not recognizable, because of the anatomic distortion, but the mass still is totally exophytic. It extends into and fills the CPA cistern mostly, may extend as far as the prepontine cistern, sometimes even beyond the basilar Fig. 19 Ependymoma, roof subtype. The tumor develops into the lingula/superior medullary velum (a–b), and therefore may easily be confused with a medulloblastoma

artery. Classically, it also fills the fourth ventricle, but it may remain completely extraventricular. It often surrounds and encases the terminal segment of the vertebral artery, the first segment of the PICA, and sometimes also the basilar artery. Cranial nerves may be stretched and whether they are infiltrated may be impossible to tell. They are not recognized if they are encased in the tumor mass. These features of the lateraltype ependymoma are usually easy to recognize, with the plastic extension of the mass that molds the cisterns. WNTgroup medulloblastomas also develop from the LRL in the lateral recess, may grow in the CPA cistern and demonstrate restricted diffusion. However, they are far less plastic than ependymomas, and even when they fill the cisterns, they do not Bcreep^ as far as ependymomas do. Roof-type ependymomas are uncommon. They develop in the cerebellum (rhombencephalic roof plate) and may therefore be easily confused with a diffusion-restricted cerebellar medulloblastoma (Fig. 19a, b).

Cerebellar pilocytic astrocytoma Cerebellar pilocytic astrocytoma (PA) is mentioned here for completeness only. Its appearance is quite specific. It never shows restricted diffusion on DWI/ADC imaging and

Childs Nerv Syst (2015) 31:1661–1676

1675

Fig. 20 Central vermian pilocytic astrocytoma. The mass develop from the cerebellum into the ventricle (a–b), and although its glial nature is easy to recognized, it may be difficult to differentiate from a glioma of the tegmentum if no CSF plane is identified around the tumor

therefore cannot be confused with either medulloblastoma or ependymoma. Sometimes however, when it is centrally located, it may be difficult to differentiate from an exophytic glioma of the ponto-medullary tegmentum (Fig. 20a, b). From our neuroradiology records of the last 15 years, 55 cases of PA were retrieved (one pilomyxoid), in addition to two gangliogliomas. A third of the PAs were solid or mostly solid, and two thirds were solid-cystic. As for the location, 53 % were on the midline, 26 % were close to the midline, and 21 % were more lateral. Associated hydrocephalus was the rule. Enhancement also was very common: only two cases (both solid) did not enhance at all. In most, the solid portion only enhanced (72 %), including 7 cases in which it suggested an associated necrosis. In 25 %, the wall of the cyst enhanced also, at least in part.

2.

3.

4.

5.

6.

7. 8. 9.

Conclusions MR imaging is recognized as the best tool to assess the structure of the posterior fossa tumors, and the MRI findings typically correlate with the pathology. Beyond this, and more and more, the tumor specificity can be identified with specific molecular markers which relate to the developmental history of the brain region from which they develop. A complete approach to the diagnosis must therefore integrate also the developmental anatomy of the brainstem and cerebellum, as tumors with an apparently similar histology but different sites of origin may have very different biologic behavior. Conflict of interest We have no conflict of interest.

10. 11.

12.

13. 14. 15.

16.

17.

References 18. 1.

Boydston WR, Sanford RA, Muhlbauer MS, et al. (1991-1992) Glioma of the tectum and periaqueductal region of the mesencephalon. Pediatr Neurosurg 17:234–238

Broniscer A, Laningham FH, Kocak M, et al. (2006) Intratumoral hemorrhage among children with newly diagnosed, diffuse brainstem glioma. Cancer 106:1364–1371 Cambronero F, Puelles L (2000) Posterocaudal nuclear relationships in the avian medulla oblongata: a fate map with quail chick chimeras. J Comp Neurol 427:522–545 Conway AE, Reddick WE, Yuan Y, et al. (2014) BOccult^ postcontrast signal enhancement in pediatric diffuse intrinsic pontine glioma is the MRI marker of angiogenesis? Neuroradiology 56(5):405–412 Curless RC, Bowen BC, Pattany PM, et al. (2002) Magnetic resonance spectroscopy in childhood brainstem tumors. Pediatr Neurol 26:374–378 Dağlioğlu E, Çataltepe O, Akalan N (2003) Tectal gliomas in children: the implications for natural history and management strategy. Pediatr Neurosurg 38:223–231 Epstein F, Wisoff J (1987) Intraaxial tumors of the cervicomedullary junction. J Neurosurg 67:483–487 Epstein FJ, Farmer JP (1993) Brain-stem glioma growth patterns. J Neurosurg 78:408–412 Gibson P, Tong Y, Robinson G, et al. (2010) Subtypes of medulloblastoma have distinct developmental origins. Nature 468:1095–1099 Gilbertson RJ, Gutmann DH (2007) Tumorigenesis in the brain: location, location, location. Cancer Res 67(12):5579–5582 Godfraind C, Kaczmarska J, Kocak M, et al. (2012) Distinct disease-risk groups in pediatric supratentorial and posterior fossa ependymomas. Acta Neuropathol 124(2):247–257 Ikezaki K, Matsushima T, Inoue T, et al. (1993) Correlation of microanatomical localization with postoperative survival in posterior fossa ependymomas. Neurosurgery 32(1):38–44 Jandial R, U H, Levy ML, Snyder EY (2008) Brain tumor stem cells and the tumor microenvironment. Neurosurg Focus 24(3-4):E27 Konno H, Yamamoto T, Iwasaki Y, et al. (1985) A case of quadrigeminal hamartoma. Acta Neuropathol 68:155–159 Kool M, Korshunov A, Remke M, et al. (2012) Molecular subgroups of medulloblastoma: an international meta-analysis of transcriptome, genetic aberrations, and clinical data on WNT, SHH, Group 3, and Group 4 medulloblastomas. Acta Neuropathol 123:473–484 Koral K, Mathis D, Gimi B, et al. (2013) Common pediatric cerebellar tumors: correlation between cell density and apparent diffusion coefficient metrics. Radiology 268(2):532–537 Krieger MD, Blüml S, McComb JG (2003) Magnetic resonance spectroscopy of atypical diffuse pontine masses. Neurosurg Focus 15(1):E5 Laprie A, Pirzkall A, Haas-Kogan DA, et al. (2005) Longitudinal multivoxel MR spectroscopy study of pediatric diffuse brainstem glioma treated with radiotherapy. Int J Radiat Oncol Biol Phys 62: 20–31

1676 19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30. 31.

32.

33.

34.

35.

36.

37.

Childs Nerv Syst (2015) 31:1661–1676 Löbel U, Sedlacik J, Sabin ND, et al. (2010) Three-dimensional susceptibility-weighted imaging and two-dimensional T2*-weighted gradient-echo imaging of intratumoral hemorrhages in pediatric diffuse intrinsic pontine glioma. Neuroradiology 52:1167–1177 Löbel U, Sedlacik J, Reddick WE, et al. (2010) Quantitative diffusion-weighted and dynamic susceptibility-weighted contrastenhanced perfusion MR imaging analysis of T2 hypointense lesion components in pediatric diffuse intrinsic pontine glioma. AJNR Am J Neuroradiol 32(2):315–322 Louis DN, Ohgaki H, Wiestler OD, et al. (2007) The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 114(2):97–109 Mack SC, Witt H, Piro RM, et al. (2014) Epigenomic alterations define lethal CIMP-positive ependymomas of infancy. Nature 506: 445–450 Marcorelles P, Fallet-Bianco C, Oury JF, et al. (2005) Fetal aqueductal glioneuronal hamartoma: a clinicopathological and physiopathological study of three cases. Clin Neuropathol 24: 155–162 Martinez S, Audreu A, Mecklenburg N, Echevarria D (2013) Cellular and molecular basis of cerebellar development. Front Neuroanat 7:18 Nakamura H, Watanabe Y (2005) Isthmus organizer and regionalization of the mesencephalon and metencephalon. Int J Dev Biol 49:231–235 Northcott PA, Korshunov A, Witt H, et al. (2011) Medulloblastoma comprises four distinct molecular variants. J Clin Oncol 29(11): 1408–1414 Northcott PA, Korshunov A, Pfister SM, Taylor MD (2012) The clinical implications of medulloblastoma subgroups. Nat Rev Neurol 8:340–351 Panigrahy A, Nelson Jr MD, Finlay JL, et al. (2008) Metabolism of diffuse intrinsic brainstem gliomas in children. Neuro-Oncology 10(1):32–44 Panigrahy A, Nelson Jr MD, Blüml S (2010) Magnetic resonance spectroscopy in pediatric neuroradiology: clinical and research applications. Pediatr Radiol 40(1):3–30 Pajtler K, Witt H, Sill M et al. The ependymoma classification. Cancer Cell 2015 (accepted). Perreault S, Ramaswamy V, Achrol AS, et al. (2014) MRI surrogates for molecular subgroups of medulloblastoma. AJNR Am J Neuroradiol 35:1263–1269 Raghunathan A, Wani K, Armstrong TS, et al. (2013) Histological predictors of outcome in ependymoma are dependent on anatomic site within the central nervous system. Brain Pathol 23:584–594 Ramaswamy V, Remke M, Bouffet E, et al. (2013) Recurrence patterns across medulloblastoma subgroups: an integrated clinical and molecular analysis. Lancet Oncol 14(12):1200–1207 Ramaswamy V, Remke M, Shih D, et al. (2014) Duration of the prediagnostic interval in medulloblastoma is subgroup dependent. Pediatr Blood Cancer 61(7):1190–1194 Ramaswamy V, Remke M, Adamski J et al. Medulloblastoma subgroup-specific outcomes in irradiated children: who are the true high-risk patients? Neuro-Oncology 2015. [Epub ahead of print] Rauscher A, Sedlacik J, Barth M, et al. (2005) Noninvasive assessment of vascular architecture and function during modulated blood oxygenation using susceptibility-weighted magnetic resonance imaging. Magn Reson Med 54:87–95 Robertson PL, Murazko KM, Brunberg JA, et al. (1995) Pediatric midbrain tumors: a benign subgroup of brainstem gliomas. Pediatr Neurosurg 22:65–73

38.

Rodriguez Gutierrez D, Awwad A, Meijer L, et al. (2014) Metrics and textural features of MRI diffusion to improve classification of pediatric posterior fossa tumors. AJNR Am J Neuroradiol 35:1009– 1015 39. Roussel MF, Hatten ME (2011) Cerebellum: development and medulloblastoma. Curr Top Dev Biol 94:235–282 40. Sanford RA, Bebin J, Smith WR (1982) Pencil glioma of the aqueduct of Sylvius. J Neurosurg 57:690–696 41. Sehgal V, Delproposto Z, Haacke EM, et al. (2005) Clinical applications of neuroimaging with susceptibility-weighted imaging. J Magn Reson Imaging 22:439–450 42. Sehgal V, Delproposto Z, Haddar D, et al. (2006) Susceptibilityweighted imaging to visualize blood products and improve tumor contrast in the study of brain masses. J Magn Reson Imaging 24: 41–51 43. Sgaier SK, Millet S, Villanueva MP, et al. (2005) Morphogenetic and cellular movements that shape the mouse cerebellum; insights from genetic fate mapping. Neuron 45(1):27–40 44. Sun B, Wang CC, Wang G (1999) MRI characteristics of midbrain tumors. Neuroradiology 41:158–162 45. Taylor MD, Poppleton H, Fuller C, et al. (2005) Radial glia cells are candidate stem cells of ependymoma. Cancer Cell 8:323–335 46. Taylor MD, Northcott PA, Korshunov A, et al. (2012) Molecular subgroups of medulloblastoma: the current consensus. Acta Neuropathol 123:465–472 47. ten Donkelaar HJ, Lammens M, Cruysberg JRM, Cremers CWJR (2006) Development and developmental disorders of the cranial nerves. In: ten Donkelaar HJ, Lammens M, Hori A (eds) Clinical Neuroembryology. Springer-Verlag, Berlin Heidelberg, pp. 274– 279 48. Ternier J, Wray A, Puget S, et al. (2006) Tectal plate lesions in children. J Neurosurg 104(6 Suppl Pediatrics):369–373 49. Tihan T, Zhou T, Holmes E, et al. (2008) The prognostic value of histological grading of posterior fossa ependymomas in children: a children’s oncology group study and a review of prognostic factors. Mod Pathol 21:165–177 50. Tümpel S, Wiedemann KR (2009) Hox genes and segmentation of the vertebrate hindbrain. Curr Top Dev Biol 88:103–137 51. U-King-Im JM, Taylor MD, Raybaud C (2010) Posterior fossa ependymomas: new radiological classification with surgical correlation. Childs Nerv Syst 26:1765–1772 52. Vandertop WP, Hoffman HJ, Drake JM, et al. (1992) Focal midbrain tumors in children. Neurosurgery 31:186–194 53. Witt H, Mack SC, Ryzhova M, et al. (2011) Delineation of two clinically and molecularly distinct subgroups of posterior fossa ependymoma. Cancer Cell 20:143–157 54. Witt H, Korshunov A, Pfister SM, Milde T (2012) Molecular approaches to ependymoma: the next step(s). Curr Opin Neurol 25(6): 745–750 55. Wullimann MF, Mueller T, Distel M, et al. (2011) The long adventurous journey of rhombic lip cells in jawed vertebrate: a comparative development analysis. Front Neuroanat 5:27 56. Xu Q, Wilkinson DG (2013) Boundary formation in the development of the vertebrate hindbrain. WIREs Dev Biol 2:735–745 57. Yeom KW, Mobley BC, Lober MR, et al. (2013) Distinctive MRI features of pediatric medulloblastoma subtypes. AJNR Am J Neuroradiol 200:895–903 58. Zhukova N, Ramaswamy V, Remke M, et al. (2013) Subgroupspecific prognostic implications of TP53 mutation in medulloblastoma. Clin Oncol 31(23):2927–2935