Historically, the cerebellum has been primarily recognized for its role in motor control; however, recent studies also suggest a significant role for this brain structure in cognitive and emotional processing. In line with this development, there has also been increased attention to its involvement in mental disorders and cognitive deficits, and a growing number of neuroimaging studies assessing cerebellar structure and function in these disorders. However, the existing neuroimaging literature on the cerebellum frequently exhibits inconsistencies, and effect sizes tend to be small. One likely reason for this is the marked heterogeneity typically seen in mental and neurological disorders, both concerning clinical symptoms and brain phenotypes. While this heterogeneity is generally acknowledged, most research designs are still limited to case-control comparisons, which effectively only assess differences between the “average patient” and the “average healthy control participant.” Moreover, since traditional parcellations of the cerebellum based on anatomical landmarks only partially overlap with more recent functional parcellations, there is no group consensus on the optimal method to quantify and represent cerebellar anatomy. Another key limitation in much existing research is the use of relatively small samples, and the out-of-sample predictive value of cerebellar anatomical alterations with respect to classifying mental and neurological illnesses remains largely unknown. Finally, the cerebellum remains relatively understudied in rare genetic syndromes associated with significantly increased risk for mental disorders.
The current PhD thesis addresses these current limitations by 1) constructing normative population models of cerebellar anatomy (akin to pediatric growth charts) based on extensive neuroimaging samples covering the human life span (n > 54k, 3-85 years), 2) estimating these models at the voxel-level, enabling integration with a wide range of cerebellar parcellations, 3) analyzing how patients with mental and neurological disorders deviate from these population reference models, 4) testing if these deviations are predictive of illness status in independent test samples, and 5) examining whether significant deviations from these normative models can be detected in genetic syndromes with increased risk for mental disorders.
This thesis reports the largest cerebellar model to date at fine-grained spatial resolution, designed to be translatable across diverse cerebellar parcellation schemes. Our findings reveal significant cerebellar anatomical alterations in individuals with schizophrenia (SZ) and autism spectrum disorder (ASD). However, predictive models trained on cerebellar deviations achieved only limited accuracy when applied to independent samples. We also observed cerebellar changes in individuals with 15q11.2 copy number variants (CNVs), a known genetic risk factor for neurodevelopmental disorders, although the effect sizes were small. Importantly, our analyses demonstrate a high degree of cerebellar heterogeneity across mental and neurological disorders as well as CNVs. Overall, this work offers insights into the cerebellar involvement in brain disorders, thereby advancing our understanding of its complex and heterogeneous architecture.