136 Morphological Clusters of Chiari I Patients With Variable Response After Suboccipital Decompression

Abstract

INTRODUCTION: Chiari malformation type I (CM-I) can cause debilitating neurologic symptoms, most notably distinctive headaches, and the only existing definitive treatment is suboccipital decompression. While CM-I is typically defined as ≥5 mm downward displacement of cerebellar tonsils, overall CM-I cerebellar morphology is highly heterogeneous, and it is uncertain if this variability represents distinct patient populations or influences clinical response to suboccipital decompression. METHODS: 186 adult CM-I patients with high-resolution T1-MRIs were retrospectively identified. Each patient’s cerebellum was non-linearly and diffeomorphically transformed to template brain, with this transformation’s inverse representing cerebellar deformation. A similarity matrix was generated between patients’ deformations using the Pearson correlation coefficient of their inverse transformations’ log Jacobian determinants within each cerebellar lobule. Spectral clustering was performed on this similarity matrix to identify patient clusters. Headache freedom after suboccipital decompression was compared between clusters by Kaplan-Meier analysis. RESULTS: Three morphological clusters of CM-I patients were identified. Cluster A is characterized by posteroinferior cerebellar fullness and platybasia; B by midline anterior-posterior compression; C solely by tonsillar descent. After suboccipital decompression, headaches were found to recur more frequently and sooner in Cluster B (p < 0.01). CONCLUSIONS: These analyses suggest that CM-I encapsulates morphologically distinct subtypes, here identified as Cluster A and Cluster B, with Cluster C representing an intermediate phenotype. These clusters vary significantly in clinical response to suboccipital decompression, which may have implications for surgical patient selection or operative technique. Future research should involve independent validation of this cluster scheme and optimization of treatment for patients in Cluster B, which is found here to have poorer surgical outcomes.