A sub-set of cases from the Iowa Neurological Patient Registry, a large database of individuals with focal brain injuries, were retrospectively examined. The study period ranged from 1994 to 2022. Available data included demographic information, diagnosis, etiology, and performance on neuropsychiatric scales. Inclusion criteria for induction into the Iowa Neurological Patient Registry were: a clear and analyzable brain lesion that can be identified on magnetic resonance imaging (MRI), age 18 years or older, and ability to provide informed consent. Exclusion criteria were: evidence of preexisting dementia, significant cognitive impairment, severe psychiatric disease, and current or recent history of heavy alcohol or drug use. Inclusion criteria for the present study were: a diagnosis of cerebellar ischemic or hemorrhagic stroke. Patients with a diagnosis of epilepsy were excluded. All participants provided informed consent. Data collection was conducted with approval from the University of Iowa Institutional Review Board.

Cerebellar vascular accidents comprise a minority of all strokes, approximately 5%. Because such cases are uncommon, all Iowa Neurological Patient Registry cases of vascular accidents affecting the cerebellum were analyzed, including participants whose damage extended into other brain regions. Within the 38-year study period, 26 cases met inclusion criteria. Limitations of this study design are further addressed in the discussion. Participant characteristics are presented in Table 1.

In our neuropsychiatric battery, cognition was assessed via the Stroop Test (Stroop, 1935), Trail Making Test (TMT; Reitan, 1956), select subtests from the Wechsler Adult Intelligence Scale 4th Edition (Wechsler, 2008), Rey Auditory Verbal Learning Test (RAVLT; Rey, n.d.), Controlled Oral Word Association Test (COWAT; Al et al., 1997), and Rey-Osterrieth Complex Figure Test (CFT; Rey and Osterrieth, 1941). Motor performance was assessed via the Lafayette Grooved Pegboard Test (GPT; Matthews and Klove, 1964). Affect was assessed via the Beck Depression Inventory—II (BDI; Beck et al., 1996), and the Beck Anxiety Inventory (BAI; Steer and Beck, 1997). Due to time constraints or participant fatigue, not all participants completed all tests. A summary of tests completed is presented in Table 2.

For comparison to healthy populations, z-scores were calculated using population means and standard deviations from testing manuals. This transformation accounts for participant age, years of education, and, where appropriate, sex. Positive z-scores indicate better performance while negative z-scores indicate worse performance, compared to the neuronormative population. TMT and GPT scores, which indicate time-to-completion, were inverted for visualization purposes such that lower scores indicate poorer performance. Mood measures (BDI and BAI) are not converted into z-scores but rather reported as categories.

Performance on neuropsychiatric measures was assessed in GraphPad Prism version 8.4.3 (GraphPad Software, San Diego, California, United States) using a one sample t-test. Multiple comparisons were corrected for using the Benjamini-Hochberg method. For evaluation of individual subjects’ performance, participants with z-scores below 2.8 (alpha = 0.05) were considered significantly impaired.

To visualize lesion locations, lesions were manually traced on native brain scans in FSL (Jenkinson et al., 2012). The native MRI was then co-registered to the MNI152 1 mm atlas using non-linear registration with Advanced Normalization Tools (Avants et al., 2008). The spatial transformation warp was applied to the native lesion mask. The anatomical accuracy of each lesion mask was reviewed in both native and MNI space and edited as needed. Lesion volumes were calculated by multiplying the number of affected volumes by voxel volume (1 mm³).

Lesion subtractions were created to visualize areas, which were uniquely affected in individuals who suffered significant impairments following a vascular accident vs. those who did not. Subtractions were performed according to methods detailed in Rudrauf et al. (2008). In brief, for each neuropsychiatric test, separate lesion overlap maps were created for individuals who showed significant impairments vs. those who did not. Each lesion overlap map was divided by its own max-overlap in order to place all maps on the same proportional scale. The proportional lesion map for unimpaired individuals was then subtracted from the proportional lesion map for impaired individuals.

Twenty-six cases of damage to the cerebellum were identified. Sixteen participants were male and 10 female. Age of vascular accident-onset ranged from 29 to 92 years with an average of 58 years. The average time between stroke-onset and neuropsychological testing was 1.37 years (SD 1.8 years). On average, participants were mildly depressed with a BDI score of 14.13, however, scores were considerably heterogeneous ranging from 0 (no depression) to 35 (severe depression). Anxiety scores were similarly heterogeneous ranging from 1 (minimal anxiety) to 31 (severe anxiety), with an average of 9 indicating mild anxiety.

85% of cases (n = 22) were diagnosed with ischemic strokes, while 15% of cases (n = 4) were diagnosed with hemorrhagic strokes. 50% of cases (n = 13) suffered right cerebellar damage, 46% (n = 12) left cerebellar damage, and 5% (one case) bilateral cerebellar damage. For cases including notes on which sub-tentorial arteries were affected (n = 7), damage was most common to the vertebral artery (57%), followed by the posterior inferior cerebellar artery (28%). There was considerable heterogeneity in cerebellar lesion size, ranging from 261 to 30,382 mm³, with an average lesion size of 6,790 mm³. Maximum lesion overlap (n = 7) occurred in right crus II, followed by left lobule VIIIa and a portion of lobule VIIIb (n = 6; Figure 2). For individual lesion maps, see Supplementary Figure 1. For assessment of patients with legions restricted to the cerebellum see Supplementary material.

The main goal of the present case-series was to evaluate the effects of strokes involving the cerebellum using a neuropsychiatric battery in the chronic (+1 year) phase following vascular accidents. Visual inspection of Figure 3 suggests that the highest average level of impairment was observed in Stroop—word (mean − 1.77; SD ± 1.45), GPT—non-dominant (mean − 1.44; SD ± 1.79), and trail making B (mean − 0.76; SD ± 1.46). See Table 3 for means and test statistics associated with each test.

Test statistics were examined for significance using an alpha of 0.05; however, no significant effects were found. Due to the exploratory nature of these analyses, we also assessed the data for marginal effects, assuming an alpha of 0.10. Compared to neuronormative controls, patients showed marginally impaired performance in Stroop word [t(9) = 3.871, p = 0.0760], Stroop color [t(9) = 3.001, p = 0.0832], RAVLT trial 1 [t(19) = 2.526, p = 0.0824], GPT—dominant [t(15) = 2.619, p = 0.0824], and GPT—non-dominant [t(13) = 3.020, p = 0.0824]. Performance on the neuropsychiatric battery did not differ by side of lesion or chronicity (data not shown).

To more closely examine scales where marginal effects were detected, we next examined individual performance within each scale. Significant impairment for an individual was defined as a z-score of −2.8 or below, representing two standard deviations below the group mean. For GPT—dominant, four patients (23%) showed significantly impaired scores (Table 4). This group included individuals with strokes in the cerebellum only (50%), and the cerebellum, cortex and sub-cortex (50%). For GPT—non-dominant, two patients (13%) showed significantly impaired performance (Table 5). This group included patients from the cerebellum & cortex lesion group (50%) and the cerebellum, cortex & sub-cortex lesion group (50%).

For RAVLT trial 1, examination of single-patient z-scores revealed that no patients were classified as significantly impaired (Table 6). This suggests that the marginal effects detected on RAVLT trial 1 result from slight impairments to most patients’ performance, as opposed to significant impairments to a small sub-group of patients.

Examination of single-subject scores for the baseline word-reading trial of the Stroop test (“Stroop—word”) reveals four patients (40%) with severe impairments (Table 7). Interestingly, all patients displaying severe impairments suffered from lesions confined to the cerebellum. Finally, examination of single-subject scores for the color naming trail of the Stroop test (“Stroop—color”) reveals one subject (10%) with significant impairment whose lesion was confined to the cerebellum (Table 8).

Lesion proportion difference maps were generated for scales where marginal impairments were detected to highlight cerebellar regions where damage may be associated with worsened performance. Results for the GPT (Figures 4A,B) suggest that there was little specificity/overlap for individuals who were significantly impaired in this task. Although lesions to crus II and lobules VII/VIII (z = −55 and − 49) were not associated with significant fine-motor impairments. Results for Stroop—word (Figure 4C) suggest overlap in the crus I and lobule VII region for individuals who were significantly impaired on this task. Because the “impaired” group consisted of zero individuals and a single individual for the RAVLT and Stroop—color tasks, respectively, lesion proportion differences maps were not generated for these tasks.

In the present data-set, significant group-level differences in GPT performance were not detected. However, 23 and 13% of patients showed significant deficits in dominant and non-dominant hand performance on this task, respectively. Given the well-documented role of the cerebellum in motor coordination (Stoodley and Schmahmann, 2010; D'Angelo, 2018), deficits on the GPT, a measure of motor function, were expected. Indeed, impairments in GPT-dominant and non-dominant hand performance have previously been reported in spinocerebellar ataxia patients (Rentiya et al., 2017), a disorder marked by the degeneration of Purkinje and granule cells. Work suggests that GPT performance is related to the integrity of cerebellar medial superior regions, which are also commonly associated with motor task performance in fMRI studies (Stoodley, 2012; Rentiya et al., 2017). Given the central role of the cerebellum in motor coordination (D'Angelo, 2018), it is interesting that not all patients displayed deficits in this task. This is likely due to the chronicity associated with the vascular damage, on average more than 1-year post-stroke, which may have allowed for compensatory processes to occur.

Impaired performance was observed on Stroop—color in a single patient, and on Stroop—word in four patients. Impairments in Stroop performance, specifically on color and word sub-tests, have been reported in cerebellar ataxia patients (Braga-Neto et al., 2012). Additionally, work suggests that patients with cerebellar lesions show impairments in reading words and sentences, making errors at the letter and the word level, a phenomenon known as cerebellar dyslexia (Moretti et al., 2002). Although the causes for this phenomenon are unclear, it may result from loss of eye-movement control, leading to binocular instability. Indeed, as the cerebellum significantly contributes to binocular alignment, cerebellar pathology often leads to nystagmus (Patel and Zee, 2015). However, it is also possible that reading difficulties associated with cerebellar damage might stem from disruption of normal communication between the cerebellum and the supra-tentorial regions implicated in linguistic processing. Indeed, tract-tracing work suggests connections between the right lateral cerebellum and the left frontal language areas (Turkeltaub et al., 2016).

Lesion subtraction results for Stroop—word suggest overlap in damage to Crus I and lobule VII in individuals who were significantly impaired on this task. This is in agreement with the larger literature suggesting that bilateral lobules VI and VII are involved in reading tasks (Stoodley, 2012). Activity in lobule VII is also reported in tasks, which tax verbal working memory. Cerebellar activation during such tasks may result from sub-vocal rehearsal, an important component of working memory (Krienen and Buckner, 2009). One interesting route for future work is to disentangle the mechanisms behind the reading deficits observed to determine if they are due to loss of articulatory control, eye movement difficulties, or cognitive dysmetria-like mechanisms (Andreasen et al., 1998). Indeed, all three mechanisms may play a role. Previous work suggests that language and reading are topographically organized within the cerebellum: while articulation engages anterior sensorimotor regions corresponding to articulatory muscles, verbal fluency/word generation engage posterolateral cognitive regions (Turkeltaub et al., 2016).

The marginal impairments observed in the present case-series on the RAVLT further suggest links between the cerebellum and language processing/production. The RAVLT is primarily a verbal test with the first trial specifically assessing attention and immediate memory. Unlike the other two tests on our neuropsychiatric battery where marginal impairments were detected, average RAVLT scores were not largely driven by a small group of significantly impaired patients. Instead, most patients showed slight impairments to their scores. This indicates that, although the association is small, cerebellar activity is important for verbal working memory, a notion supported by the broader literature.

It is worth mentioning two tests where impairments were surprisingly not detected in our cohort: the CFT and the TMT. The CFT assesses visuo-constructional ability and visual memory. Drawing complex figures relies on fine-motor coordination, visuospatial perception, spatial orientation, non-verbal memory, planning, and organization (Zhang et al., 2021). The involvement of the cerebellum in visuospatial abilities has been evidenced in numerous studies (Garcia et al., 2020; Zhang et al., 2021). Work suggests that individuals with cerebellar damage show worsened performance on tasks that involve mental imagery, mental object rotation, or visuocontructive demands (Garcia et al., 2020). Work by Olivito et al. (2018) found significant correlations between visuospatial performance impairment and gray matter loss in lobules V, VIIb, VIIIa, Crus I, Crus II, and the vermis. Additionally, our study did not confirm previous reports that patients with cerebellar damage show impairments on the TMT (Braga-Neto et al., 2012; Rentiya et al., 2017). The TMT is a robust assessment of visual search speed, visual attention, task switching, and mental flexibility. Work suggests that TMT performance-difficulties are associated with gray matter degeneration in the inferior hemispheric region of lobules VIIb and VIIIa.

An assessment of the broader literature suggests significant disagreement between works describing the effects of cerebellar damage, likely resulting from heterogeneous clinical presentations. A significant contributor to this heterogeneity is differences in etiology between works. Indeed, different presentations are expected, for example, between ataxias, which selectively damage specific cell types vs. strokes, which damage all cell types and fibers of passage. Lesion chronicity is another important contributor to symptom heterogeneity. While acute symptoms of cerebellar vascular accidents may be pronounced (Edlow et al., 2008); compensatory mechanisms can lead to a positive prognosis in the chronic phase. In the present work specifically, where the mean chronicity was 1.3 years, compensatory mechanisms may have taken place resulting in a masking of the association between vascular accident location and resulting symptoms. Finally, the heterogeneity of symptoms described in the literature might also be indicative of the wide-ranging domains in which the cerebellum is involved, including motor, cognitive and affective.

In addition to the caveats addressed above, our retrospective clinical case-series design also presents significant limitations. First, race/ethnicity were largely heterogeneous within our study population consisting almost entirely of Caucasian and non-Hispanic patients. This is a particularly significant caveat in the case of work relating to stroke, as studies suggest that factors relating to racial/ethnic disparities significantly impact outcomes (Skolarus et al., 2020). Thus, outside of Caucasian/non-Hispanic individuals we do not know how the characteristics we identify would extrapolate to the larger population. Additionally, although this design allowed for the assessment of rare occurrences, the number of samples was too small to identify a true association. Indeed, in our study period of 38 years, only eight cases consisting of damage restricted to the cerebellum were available. Cases including damage to other brain regions were thus included in our sample, introducing a potential none-differential misclassification bias, which could artificially deflate or inflate the associations observed in this study. Thus, it is prudent to interpret the results of the present case-series as revealing areas where future studies may focus.