Tumor Location Definition

Tumor location was determined by 2 attending radiologists based on the area with the largest tumor mass; both attending radiologists had to reach a unanimous decision regarding the tumor location. Tumor sites were grouped into 3 categories—namely traditional midline structures, which included the thalamus, midbrain, pons, medulla oblongata, and spinal cord; nontraditional midline structures, which included the cerebellar vermis, cerebellar hemisphere, corpus callosum, and hypothalamus; and nonmidline structures, which included the cerebral hemisphere. A cerebral hemispheric lesion referred to any lesion located entirely within the cerebral hemisphere; large and irregular hemispheric lesions were documented as thalamic lesions if they were partially connected to the thalamus. A corpus callosal lesion was defined as any lesion that grew either within the corpus callosum or in the periphery without any involvement of the traditional midline structures. A whole-brain type lesion was defined as the simultaneous involvement of the cerebral hemisphere with either the brainstem, the thalamus, or both.

Contrast Enhancement and Other MRI Characteristics

Lesions were classified visually into 4 categories based on the degree of enhancement on MRI: 1) no enhancement; 2) partial enhancement (enhanced foci accounting for < 50% of the whole lesion); 3) diffuse enhancement (enhanced foci accounting for ≥ 50% of the whole lesion); and 4) irregular peripheral enhancement.

Aside from enhancement features, other imaging findings described herein included hemorrhage, edema, and diffusion restriction on diffusion-weighted imaging (DWI). Hemorrhage was defined as areas with hyperintense signals on T1-weighted images. Peritumoral edema was defined as nonenhancing regions surrounding the tumor with distinct, homogeneous hyperintensity on T2-weighted and FLAIR sequences. The degree of diffusion restriction was classified visually into unremarkable, mild, moderate, and severe. A lesion had unremarkable diffusion restriction if no hyperintensity was observed; mild diffusion restriction if the signal was slightly higher than in normal brain parenchyma; moderate if hyperintense but lower than hyperacute infarction; and severe if higher than hyperacute infarction.

Demographic Data

A total of 75 cases were identified, including 9 in children or adolescents and 66 in adults. Because the present study aimed to describe the imaging characteristics of glioma, H3 K27M-mutant, in adults, the 9 cases in children or adolescents (4 in the thalamus, 5 in the brainstem) were excluded. The remaining 66 cases in adults had a male/female ratio of 1.5:1 (40 men, 26 women). The age distribution of the patients included herein was such that 64 of the 66 patients were between 20 and 60 years of age and the remaining 2 patients were > 60 years of age (Fig. 1).

Bar graph showing age and sex distribution. A total of 66 patients (40 men, 26 women) were included in the study. The oldest patient was 78 years old at the time of diagnosis; 2 patients were > 60 years of age, and the remaining patients were within the 20- to 60-year age group. Figure is available in color online only.

Anatomical Locations

Of the 66 cases, the imaging characteristics of 61 cases were analyzed using pre- and postcontrast MR images, whereas the remaining 5 cases were analyzed using only T2-weighted images. Follow-up MR images were only available in a few cases. The lesion sites were distributed as follows (Fig. 2): 38 cases in the thalamus (18 on the right side, 11 on the left, and 9 on both sides); 6 cases all in the dorsal aspect of the brainstem (2 in the pons, 4 in the medulla oblongata), with 2 of the cases extending longitudinally to involve the cerebellum (Fig. 3A, Fig. 4E and F) and another 2 cases extending superiorly to involve the thalamus; 8 cases involving the whole brain (Fig. 3B); 3 cases in the genu of the corpus callosum (Fig. 4A and B), with 1 case extending laterally to involve the temporal lobe and another the left frontal lobe; 1 case in the hypothalamus (Fig. 4C and D); 4 cases in the spinal cord (1 each in the cervical, cervicothoracic, thoracic, and lumbar cord); and 2 cases in the cerebral hemisphere (1 each in the right temporal lobe and left occipital lobe) (Fig. 4G and H). Spinal lesions found in the cervical, cervicothoracic, and thoracic cords spanned more than 3 vertebral segments (i.e., long-segment lesions); the lesion in the lumbar spinal cord appeared as an enlargement of a short spinal segment in the region of the conus medullaris.

Diagram of tumor site distribution. The tumor sites observed in the present study included the thalamus, bilateral thalami, pons, medulla oblongata, cerebral hemispheres, infratentorial structures, multiple sites of brainstem involvement along the longitudinal axis, simultaneous involvement of the thalamus, pons, and medulla oblongata, spinal cord, and whole brain. Infratentorial structures included the brainstem and cerebellar vermis/cerebellar hemisphere. One case of bilateral thalamic lesion also extended to involve the hypothalamus. Figure is available in color online only.

H3 K27M-mutant gliomas with extensive involvement. Brain MR images showed lesions of various enhancement degrees ranging from nonenhancing to enhancing. Lesions were seen in the infratentorial regions involving the brainstem, cerebellum, and eighth cranial nerve in a 26-year-old man (A) and pons, medulla oblongata, cerebellar vermis, bilateral thalamus, and left frontal lobe in a 37-year-old woman (whole-brain type lesion) (B). C = contrast.

H3 K27M-mutant gliomas in nontraditional midline or nonmidline sites. MR images showed lesions in the corpus callosum in a 25-year-old man (A and B); hypothalamus in a 58-year-old man (C and D); cerebellar vermis in a 39-year-old woman (E and F); and right temporal lobe in a 20-year-old woman (G and H).

MRI Findings

Sixty-one cases were analyzed using contrast-enhanced T1-weighted MR images; of these cases, 11 demonstrated no enhancement, 25 showed partial enhancement, and the remaining 25 showed either diffuse or irregular peripheral enhancement. The main enhancement patterns observed in this study included those described as being ring-like, patchy, homogeneous, and cystic with nodules in appearance. Unlike true glioblastoma lesions, which typically present with uneven wall thickness and mixed capsule signals on contrast-enhanced T1-weighted MR images, most lesions with irregular peripheral enhancement had thin walls with relatively uniform thickness, along with capsules that demonstrated relatively homogeneous, CSF-like signals (Fig. 5). Both no enhancement and irregular peripheral enhancement were observed in long-segment spinal lesions. Most parts of the temporal lobe tumor showed solid, heterogeneous enhancement with a small amount of peritumoral edema; the tumor extended medially to the lateral margin of the right thalamus without infiltrating the structure. The occipital lobe lesion was extensive with irregular peripheral enhancement and appeared to be compressing the brainstem.

The variable degrees and patterns of tumor contrast enhancement observed in the study. MR images of the brain showed a series of right thalamic tumors with different degrees and patterns of enhancement, including tumors with no enhancement in a 28-year-old man (A and B); partial enhancement in a 33-year-old man (C and D); diffuse and substantially obvious enhancement in a 33-year-old man (E and F); diffuse and obvious heterogeneous enhancement in a 27-year-old woman (G and H); and irregular peripheral enhancing lesion in a 26-year-old man (I). Brain MR image of a 25-year-old woman showed an enhancing tumor with cystic change in the left thalamus extending to the basal ganglia (J).

Peritumoral edema was observed in 10 cases, whereas hemorrhage was only observed in 1 case located in the corpus callosum. Peritumoral edema was found adjacent to the lesions and manifested on MRI as obvious ring enhancement. All of the lesions appeared to have either moderate or unremarkable diffusion restriction on DWI, with no lesion demonstrating obvious restricted diffusion.

Tumor Dissemination and Spread

In the present study, leptomeningeal dissemination was observed in 7 cases of intracranial and 1 case of conus medullaris lesion. Nodular enhancement of the brain parenchyma was observed in 1 case (Fig. 6). Leptomeningeal dissemination was not limited to cases with enhancing primary lesions and could also be found in nonenhancing cases. Three cases of subependymal spread were observed, in which extensive growth along the lateral ventricles resulted in the irregular thickening of ventricular walls and abnormal signals on T2/FLAIR.

An example of CSF dissemination in a 41-year-old woman presenting with a tumor in the conus medullaris. Spine MR images showed a short segment of abnormal enlargement in the conus medullaris along with diffuse leptomeningeal enhancement (A–C). Brain MR image also showed 2 enhancing nodules in the right hippocampal head and cerebral peduncle region, presumably via CSF dissemination (D).

Rates of Tumor Progression

Two cases documented herein were particularly notable for their progression rates. The first case described a slow-growing, left thalamic lesion in a 39-year-old man, which was observed radiographically for 34 months before the patient underwent surgery and received the diagnosis of diffuse midline glioma, H3 K27M-mutant (Fig. 7).³ The second case described a rapidly progressive, nonenhancing lesion in the dorsal pons of a 43-year-old man, which enlarged and progressed into a lesion with irregular peripheral enhancement within 7 months of initial presentation (Fig. 8). Both patients were subsequently lost to follow-up.

Slow-growing thalamic glioma in a 39-year-old man. T2-weighted MR image revealed a left thalamic lesion at initial presentation (A). A series of MR images taken in the months following the initial scan showed the tumor’s slow progression and the invasion of the contralateral thalamus 18 months later (B–E). T1-weighted contrast-enhanced MR image obtained 34 months after initial presentation showed a nonenhancing lesion (F). Reprinted from World Neurosurg, Vol. 127, Chanchotisatien A, Pan J, Du Z, et al., Slow-growing thalamic glioma with histone H3 lysine 27-to-methionine mutation: 3-year follow-up before surgical intervention, 266–268, Copyright (2019), with permission from Elsevier.

A rapidly progressive tumor in a 43-year-old man. Initial MR images showed a nonenhancing tumor in the dorsal part of the brainstem (A and B). The MR images obtained 7 months later revealed an enlarged, enhancing tumor (C and D).

Midline Location

We found that unlike pediatric cases of H3 K27M-mutant gliomas,^1,17 adult cases tend to occur most frequently in the thalamus (58%) rather than the brainstem (9%), which also corresponds to the study carried out by Schreck et al.¹⁶ Of all thalamic cases, right-sided lesions were most common (47%), but bilateral thalamic lesions were also frequently seen, accounting for 24% of all thalamic cases. Thalamic lesions tend to grow diffusely by expanding anteroposteriorly, lateromedially, and rostrocaudally, thereby occupying larger areas of the brain compared to other locations in this cohort.

Another interesting difference we observed is the unprecedented occurrence of lesions in the dorsal aspect of the brainstem in all 6 cases of pure brainstem tumors.^4,8,17 Pure brainstem lesions were also found to be partially exophytic. In our cohort, brainstem lesions were predominantly solid with mostly homogeneous signals, and enhancement and edema were rarely seen. Because all diffuse midline gliomas harboring H3 K27M mutations are by definition WHO grade IV tumors, the lack of consistent tumor enhancement suggests that the degree of enhancement is not representative of tumor grade, but rather that the histopathological heterogeneity of these tumors determines tumor grade.⁴ In addition, brainstem lesions may also extend longitudinally to involve the cerebellum. Although no pure cerebellar lesion was found in the present study, there have been several reported cases of adult H3 K27M-mutant gliomas presenting in the hindbrain-derived cerebellar hemispheres in the literature.^6,9,14,18 The presence of fetal hindbrain-derived neural progenitor cells in H3 K27M-mutant gliomas in recent reports gives reason to speculate that such malignancy may also occur in the hindbrain-derived cerebellar vermis.¹⁸

Nonmidline Location

The presentation of H3 K27M-mutant glioma in a nonmidline (i.e., hemispheric) location is rare.^5,6,8,17 Only 2 cases of hemispheric lesions were seen in the present study. Whereas most documented hemispheric lesions were infiltrating and involved other midline structures,^5,6,8,15,17 both cases included in this study were purely hemispheric without any midline involvement (Fig. 4G and H). MR findings resembled those of glioblastoma, in that the lesions were solid with heterogeneous enhancement and surrounded by small areas of peritumoral edema. However, unlike the typically aggressive presentation of hemispheric glioblastoma, follow-up scans obtained 2 years after a partial tumor resection in 1 case revealed no obvious progression of the residual enhancing lesions. Hemispheric lesions in both pediatric and adult cases have been reported in only a few studies to date.^5,6,8,15,17 One study reported 61 cases of pediatric high-grade glioma, among them 38 cases of H3 K27M-mutant glioma, 4 of which were located in the cerebral hemisphere.⁵ Another study analyzed 28 cases of circumscribed glioma with H3 K27M mutations collected from The Cancer Genome Atlas between 2012 and 2017, among which 1 case of hemispheric lesion was included.^6,15

Whole-Brain Type Lesion

Whole-brain type lesions accounted for 12% of all cases and presented similarly to the formerly distinct entity gliomatosis cerebri, an extensively infiltrating glioma involving 3 or more contiguous lobes in the brain, but also with concurrent involvement of 1 or more traditional midline structures. MRI showed a wide range of abnormal signals in either one or both cerebral hemispheres, along with the typical imaging features of midline glioma. In the present study, hemispheric lesions of the whole-brain type demonstrated irregular peripheral enhancement, and infratentorial and thalamic lesions enhanced poorly with contrast. Previous studies revealed that apart from H3 K27M mutations, the dissemination of H3 K27M-mutant glioma to different sites is also accompanied by distinct essential accessory mutations, possibly explaining the difference in the degree of enhancement observed in supratentorial and infratentorial lesions.¹¹ Due to their similar presentations, we recommend that tumors presenting with gliomatosis cerebri–like growth patterns concurrently involving the brainstem or thalamus be tested for H3 K27M mutation when possible.

Tumor Dissemination and Spread

CSF dissemination, as demonstrated on MRI by either diffuse or nodular leptomeningeal enhancement, was detected on the initial scans in 8 cases (12%). CSF dissemination appeared to be independent of the degree of lesion enhancement, as exemplified by the case of a nonenhancing conus medullaris lesion presenting with both leptomeningeal and intracranial dissemination via the CSF route. Subependymal spread can result in irregular ventricular wall thickening and signal abnormalities on MRI.

Contrast Enhancement

Interestingly, most lesions in our study were solid with relatively homogeneous signals on plain MRI. Contrast-enhanced MRI, however, revealed various degrees and patterns of lesion enhancement ranging from nonenhancing to irregular peripheral enhancement. Histopathological studies of both biopsy and autopsy specimens in the literature have also identified a variety of findings ranging from low-grade to high-grade lesions and circumscribed to diffuse growth patterns.^2,15,17 The highly variable presentation on MRI is reflective of the different gross pathological changes, which in turn are influenced by the molecular and histopathological heterogeneity associated with this entity.¹¹ Although H3 K27M-mutant gliomas are known to be heterogeneous, a recent study carried out on autopsied brains has also discovered the presence of temporal and spatial homogeneity in certain molecular and pathological changes in gliomas that were not only located in different areas of the brain but also presented with different morphological features,¹¹ raising the possibility of targeted therapies being developed against such tumors. Although the idea may seem ambitious, we speculate that the presence of such homogeneity could also potentially allow biopsy sites to be preferentially selected for their safety when molecular pathology testing is available.

The limitations of this study include the lack of detailed information regarding the patients’ clinical presentations and treatment modalities, along with the lack of statistical data on patient survival. In addition, no volumetric measurement was made when defining the degree of lesion enhancement. Direct measurement of the signal intensity of DWI was also not performed. The lack of numerical measurements of the aforementioned variables may lead to discrepancies and difficulties in reproducing data for future works. Postoperative MR images were only available in a few cases, rendering it impossible to generate any statistical data on tumor progression rates. However, we included 2 cases of dissimilar progression rates as reminders that much is still unknown about the H3 K27M mutation and the extent of its impact on the disease course.³ Future studies with more comprehensive clinical data are warranted to explore the other aspects of H3 K27M-mutant gliomas not discussed in this paper, including management and outcome. We are currently in the process of gathering more clinical data and following up on all adult cases, and we intend to perform a thorough analysis of these clinical variables in a follow-up study. Nevertheless, the present study has successfully illustrated a spectrum of imaging characteristics associated with this type of malignancy in the adult patient, which we believe will be useful in guiding future works.