Abstract
OBJECTIVE
Chiari malformation type I (CM-I) is the most common structural hindbrain defect in humans and causes debilitating headaches and other neurological symptoms frequently requiring surgery. Although patients’ brains have heterogeneous shapes, diagnosis is based on a single radiographic measure with only 1 available pathology-directed treatment. The authors sought to determine whether there are distinct morphological subtypes of CM-I and to determine whether patients with different subtypes differ in clinical response to neurosurgical decompression.
METHODS
For this cohort study, 186 adult patients with CM-I (152 females, 34 males; mean age 40.6 years) with high-resolution T1-weighted MR images were retrospectively identified from the records of 2 large-volume tertiary care centers. Cerebellar morphology was measured by nonlinearly transforming T1-weighted MRI data to a template brain to generate cerebellar deformation maps, and spectral clustering was performed to identify morphological patient subtypes. Headache freedom after posterior fossa decompression (PFD) was compared between morphological clusters using Kaplan-Meier analysis.
RESULTS
Three distinct morphological subtypes of CM-I were identified: cluster A was characterized by posteroinferior cerebellar fullness and platybasia; cluster B, by midline anteroposterior compression; and cluster C, primarily by tonsillar descent. Similar clusters were generated after random partitioning of patients, and cluster assignments using partitioned data and the full data set were significantly associated (p
CONCLUSIONS
These analyses suggest that there are different subtypes of CM-I in patients, with distinct cerebellar shapes that respond differently to PFD, indicating that global cerebellar morphology may inform more personalized evaluation and treatment of CM-I. Future research should involve independent validation of this cluster scheme, identification of genetic and other etiological factors responsible for phenotypic heterogeneity, and optimization of treatment for patients with cluster B morphology, which was observed here to be associated with less effective surgical outcomes. Broadly, implementation of automated statistical shape analysis may represent an effective method to identify clinically meaningful subgroups of diseases characterized by structural deformations.