A data driven approach for soft tissue biomarker identification linked to Chiari-like malformation and syringomyelia.

Abstract

Canine Chiari-like malformation (CM) is a neuroanatomical condition associated with conformational change of the cranium, craniocervical junction and neuroparenchyma, resulting in pain (Chiari associated pain or CM-P) and the development of syringomyelia (SM). The associated neuro-disability in affected individuals compromises quality of life. CM is characterized by overcrowding of the brain and cervical spinal cord and is predisposed by skull-base shortening and miniaturization with brachycephalic toy dogs overwhelmingly represented. Magnetic resonance imaging (MRI) is conventionally used for diagnosis; however, CM is complex and ubiquitous in some dog breeds so that diagnosis of CM-P relies on a combination of clinical signs, MRI, and elimination of other causes of pain. This research aimed to identify cranial and spinal pathologies and neural morphologies linked to CM-P and SM in dogs using MRI scans and machine learning with the aim of identifying novel data driven biomarkers which could confirm CM-P and identify dogs at risk of developing SM. The methodology identified four regions of interest as having robust discrimination for CM-P, with 89% sensitivity and 76% specificity. A set of morphological features linked to CM-P were identified. Four regions of interest were also identified as having robust discrimination for SM, with 84% sensitivity and 80% specificity. Overall, these findings shed light on the distinct morphologies related to CM-P and SM, offering the potential for more accurate and objective diagnoses in affected dogs using MRI. These results contribute to the further understanding of the complex pathologies associated with CM and SM in brachycephalic toy pure and mixed breed dogs and support the potential utility of data-driven techniques for advancing our knowledge of these debilitating conditions.