Neuroanatomy of Cerebellar Mutism Syndrome: The Role of Lesion Location

Approximately 25% of pediatric patients who undergo cerebellar tumor resection develop cerebellar mutism syndrome (CMS). Our group recently showed that damage to the cerebellar deep nuclei and superior cerebellar peduncles, which we refer to as the cerebellar outflow pathway, is associated with increased risk of CMS. Here, we tested whether these findings replicate in an independent cohort. We evaluated the relationship between lesion location and the development of CMS in an observational study of 56 pediatric patients who underwent cerebellar tumor resection. We hypothesized that individuals that developed CMS after surgery (CMS+), relative to those that did not (CMS−) would have lesions that preferentially intersected with: 1) the cerebellar outflow pathway, and 2) a previously generated ‘lesion-symptom map’ of CMS. Analyses were conducted in accordance with pre-registered hypotheses and analytic methods (https://osf.io/r8yjv/). We found supporting evidence for both hypotheses. Compared with CMS− patients, CMS + patients (n = 10) had lesions with greater overlap with the cerebellar outflow pathway (Cohen’s d = .73, p = .05), and the CMS lesion-symptom map (Cohen’s d = 1.1, p = .004). These results strengthen the association of lesion location with risk of developing CMS and demonstrate generalizability across cohorts. These findings may help to inform the optimal surgical approach to pediatric cerebellar tumors.


Introduction
The odds of survival from pediatric brain tumors have steadily improved in recent decades [1], escalating the importance of understanding and preventing long-term treatment-related adverse effects. Approximately 25% of children undergoing cerebellar tumor resection will experience a postoperative syndrome characterized by emotional lability, executive dysfunction, and language de cits [2][3][4][5][6][7][8]. This constellation of symptoms is commonly referred to as cerebellar mutism syndrome (CMS) [9], posterior fossa syndrome [6], or cerebellar cognitive affective syndrome +/-cerebellar mutism [6]; we will use the term CMS for this paper. The duration and severity of this syndrome is variable, but importantly, patients who develop CMS typically have worse long-term cognitive outcomes [4,6,[10][11][12], highlighting the importance of this syndrome as a potential harbinger of long-term impairment.
The pathophysiology of CMS is not fully understood. Recent work by our group has demonstrated the importance of lesion location. We hypothesized that damage to the cerebellar out ow pathway, an efferent pathway passing from the deep cerebellar nuclei through the superior cerebellar peduncles to the thalamus, is associated with the development of CMS in pediatric patients [7]. This hypothesis was based on the anatomical organization of cerebro-cerebellar communication, which has an anatomically constrained bottleneck whereby even relatively small lesions involving the cerebellar out ow pathway can disrupt communication of the cerebellum with a wide array of brainstem and forebrain regions. We found support for this hypothesis from both theory-driven and data-driven analyses [7].
Identifying lesion sites associated with a higher risk of CMS has the potential to inform clinical practice. It is possible that more personalized prognostic information could be provided to patients and their families about their risk for CMS based on the tumor location relative to the cerebellar out ow pathway. It is also possible that once critical anatomical regions associated with CMS are well established, they could be identi ed in advance of the tumor resection and used to inform image guided surgery to avoid those regions when possible. However, an important prerequisite to the development of any clinical application is demonstrating that lesion location is a reliable marker of risk for CMS that generalizes to other independent cohorts. This was the primary objective of the current study. Speci cally, we aimed to evaluate whether the same anatomical regions associated with higher risk of CMS in our previous study would replicate in an independent cohort. Our anatomical hypotheses and the analytic strategy were preregistered with the Open Science Framework (https://osf.io/r8yjv/). In a cohort of 56 patients who underwent cerebellar tumor resection, we hypothesized that patients who developed CMS (CMS+) would have damage to the cerebellar out ow pathway to a greater extent than individuals who did not develop CMS (CMS-). Similarly, we hypothesized that CMS + individuals would have lesions that overlapped with a lesion-symptom 'map' of CMS previously derived from a sample of 195 patients [7].

Methods
We analyzed clinical outcome and imaging data in patients with cerebellar tumor resection surgery (n = 56) from two sites: the University of Iowa (n = 9) and Great Ormond Street Children's Hospital in London (n = 47). This study was approved by the Institutional Review Board and ethical standards committee prior to conducting this retrospective study. We included patients under the age of 21 with a diagnosis of a cerebellar tumor that had a surgical resection and follow-up imaging to show the tumor resection cavity. This study focused on a pediatric population since CMS is rare in adult patients [11]. Each patient also had clinical assessments by their treatment teams to determine whether they met criteria for CMS. CMS was de ned by criteria outlined previously: post-surgical onset of reduced speech/mutism and emotional lability [7]. Additional common features included motor dysfunction or hypotonia. Postsurgical structural neuroimaging scans performed one month or more after the surgery were selected whenever available (see Supplementary Table 1 for exact timing per participant). All scans were reviewed in advance of the analysis and only included if they were of su cient quality to clearly observe the borders of the post-surgical resection cavity. Scans were performed for clinical indications, so scanners and MRI sequence varied between patients, but all were reviewed and determined to be of su cient quality for study inclusion. The pons, medulla, brainstem, and cerebellum were isolated from the rest of the brain using the "isolate" function in the SUIT toolbox (https://www.diedrichsenlab.org/imaging/suit_function.htm). The same was done for the MNI152 atlas. The lesion mask was transformed to the MNI152 1mm brain atlas using nonlinear transformation with cost-function masking [13]. The anatomical accuracy of the lesion tracing and the transformation to MNI space was con rmed by a neurologist (A.D.B.) blinded to CMS status.
Our rst hypothesis was that CMS + patients would have greater disruption to the cerebellar out ow pathway. The region of interest (ROI) that de nes the cerebellar out ow pathway was the same as previously described [7] and is displayed in Fig. 1a. Brie y, it was produced by combining the cerebellar deep nuclei atlas (Spatially Unbiased Infra-Tentorial Template, diedrichsenlab.org/imaging/propatlas.htm) with a mask of the superior cerebellar peduncles de ned from a probabilistic atlas of these tracts [7]. The extent to which each individual patient's lesion location, or lesion 'mask', intersected with the binary cerebellar out ow pathway ROI was quanti ed and referred to as cerebellar out ow pathway 'lesion load'. The percentage of voxels in each slice of the cerebellar out ow pathway that intersected with the lesion mask was calculated. This required rst separating the cerebellar out ow pathway ROI by 1mm oblique (17°) coronal slices perpendicular to the superior cerebellar peduncles. The slice of the cerebellar out ow pathway with the highest percentage of overlap between the lesion mask and ROI was used as the value of cerebellar out ow pathway 'lesion load' for that patient. A lesion intersecting all cerebellar out ow voxels, in a given slice, would have a value of 100% and a lesion that entirely spares this out ow pathway would have a value of 0%. The lesion load values of the CMS + and CMS-groups were compared using an independent samples t-test to evaluate the one-tailed hypothesis that higher cerebellar out ow pathway lesion load will be seen in the CMS + group compared with CMS-group.
This study's second main hypothesis was similar in design to the rst but used a data-driven a priori ROI in place of the cerebellar out ow pathway ROI. We used the lesion-symptom map that was generated by Albazron and colleagues [7] using lesion and outcome data from 195 pediatric patients, displayed in Fig. 1b. The multivariate lesion-symptom mapping was performed using the R package LESYMAP [14].
LESYMAP uses sparse canonical correlation analysis for neuroimaging (SCCAN) to associate loci of brain damage with CMS status as a binary outcome. A within-sample cross-validation is performed with mapping in 75% of the patients to predict the CMS status of the remaining 25%. This determines the optimal sparseness value with the highest cross-validation correlation between the measured and predicted score [7,14]. The resulting lesion-symptom map identi ed cerebellar regions statistically associated with severe post-operative cognitive and affective symptoms, which were referred to as cerebellar cognitive affective syndrome in that study, but also met the criteria for CMS as used in the current analysis. This statistical map showed localization to the cerebellar out ow pathway, speci cally the fastigial nuclei, interposed and medial dentate nuclei, superior cerebellar peduncles and also regions outside of the cerebellar out ow pathway, including lobules IX and X of the vermis [7]. Using this lesionsymptom map as an a priori ROI we tested our second hypothesis that CMS + patients will have a higher lesion-symptom map lesion load than patients without CMS. The product of the weighted matrix of voxel values representing the lesion-symptom map, eigenvalue, and binary lesion mask of each patient in this study's cohort are standardized to calculate the lesion-symptom map lesion load. We compared the lesion-symptom map lesion load between the CMS + and CMS-groups, again using a one-tailed independent samples t-test. A lesion-symptom mapping analysis was also performed in this cohort using the same approach as Albazron and colleagues [7,14].

Results
A total of 56 pediatric patients who underwent posterior fossa tumor resections met inclusion criteria for the study. Ten patients had CMS (17.9%). The average age of the sample is 6.2 ± 3.7 years, median 6.0 years, range 5 months -14 years with various types of tumors (24 pilocytic astrocytomas, 24 medulloblastomas, 3 atypical teratoid rhabdoid tumors, 2 gangliomas, 1 ependymoma ,1 hemangioblastoma, and 1 high-grade glioma; Supplementary Table 1). Children with CMS were younger (4.4 years old ± 2.4 years, compared to CMS-6.9 years old ± 4.2 years; t(21)=-2.59, p = .016) and had similar lesion volume (6076 mm 3 ± 4891 mm 3 vs. 9831 mm 3 ± 10742 mm 3 ; t(30)=-1.7, p = .10). Medulloblastoma was the tumor type in 70% of CMS + patients and 32.6% of the CMS-cohort. The lesions of all CMS + patients crossed the midline, while 65% of CMS-lesions crossed the midline. The lesion masks from each patient were overlapped to show the spatial distribution of the lesions in the entire sample (Fig. 2a) and split by CMS status. CMS + individuals (n = 10) are shown in Fig. 2b and CMSindividuals (n = 46) are shown in Fig. 2c. A proportional subtraction map of CMS + minus CMS-lesion masks is displayed in Fig. 2d [15]. The lesion-symptom map in Fig. 2e shows that damage at MNI coordinate 1-47 -25 in the cerebellar vermis is most associated with developing CMS after tumor resection (r = .481, p < .001).
To test our a priori hypothesis that damage to the cerebellar out ow pathway increases the risk of developing CMS, cerebellar out ow pathway lesion load values were calculated for patients with and without CMS. The cerebellar out ow pathway lesion load was higher in the CMS + group relative to the CMS-group (37 ± 30% vs. 19 ± 24%, respectively; t(11) = 1.8, p = .050; Cohen's d = 0.73). Notably, the CMS rate was observed to increase in accordance with greater lesion involvement of the cerebellar out ow pathway (Fig. 3). We evaluated our second hypothesis that CMS + patients would have a greater lesionsymptom map lesion load when compared to CMS-patients. As hypothesized, CMS + patients had a greater lesion-symptom map lesion load than CMS-patients (.018 ± .010 vs. .009 ± .011, respectively; t(13) = 3.1, p = .004; Cohen's d = 1.11).

Discussion
This study evaluates lesion location in relation to developing cerebellar mutism syndrome after cerebellar tumor resection in pediatric patients. Our results support prior work in showing that lesion location is associated with risk of CMS. We [5][6][7][16][17][18][19][20][21][22] tested two pre-registered hypotheses and found supporting evidence for both. Individuals that developed CMS had lesions with greater overlap with the cerebellar out ow pathway than those that did not develop CMS. In addition, patients with CMS had lesions that overlapped to a greater extent with a lesion-symptom map of CMS derived from an independent sample [7]. Surgical resection cavities in the cerebellar vermis were most associated with CMS in the lesionsymptom mapping analysis, which is consistent with prior work. Taken together, these ndings emphasize that lesion location is useful in predicting the development of CMS in a way that generalizes across cohorts. This information may help to inform ongoing efforts to reduce the likelihood that a patient will develop CMS post-surgery, and to provide more accurate education to patients and families about the risks for developing CMS based on the precise location of the tumor There are limitations of this study. First, CMS was diagnosed by the treating clinicians without standardized assessments of behavioral de cits. It is possible that symptom-speci c quantitative assessments may further clarify unique anatomical associations with speci c symptoms. Other factors that likely in uence the development of CMS, like post-surgical treatment plan, surgical approach, edema, hydrocephalus, and premorbid cognitive abilities were not evaluated here. The observation that lesion location signi cantly relates to CMS outcome without accounting for these other variables supports the robust effect of lesion location, but more sophisticated models that take these additional factors into account are likely to explain additional variance regarding CMS risk.

Conclusion
In closing, this study provides additional evidence that damage to critical regions of the cerebellum and its out ow pathway is associated with increased risk of a child developing CMS after cerebellar tumor resection. Further work in this line of research could be used to inform the surgical approach to pediatric cerebellar tumor resection. For instance, critical anatomical regions that, when resected, are associated with increased CMS risk could be displayed as an overlay onto a patient's pre-operative MRI scan so that surgeons could design a minimally invasive MRI-guided approach that minimizes damage to these regions. Figures Figure 1 a&b 1a shows the cerebellar out ow pathway in red. 1b shows the lesion-symptom map from Albazron and colleague's 2019 study used to generate lesion-symptom map lesion load Figure 2 a -e 2a shows the lesion overlap of all patients included in this study (n=56). The region of maximum lesion overlap (n=29 of 56) was the right vermian lobule IX (MNI coordinates 4 -53 -34). 2b shows the CMS+ lesion overlap (n=10) with the maximum overlap at the right vermian lobule IX (9 of 10 lesions, MNI 3 -54 -34). 2c displays the CMS-lesion overlap with peak overlap of 21 of 46 lesions in the right vermian lobule VIII at MNI coordinate 2 -61 -36. 2d shows the proportional subtraction map with a regional peak in the anterior vermis at MNI coordinate 3 -47 -26. The lesion-symptom map in 2e also supports that lesions to the anterior vermis are associated with CMS (r=.481, p<.001, peak MNI coordinate 1 -47 -25)