The cerebellum contains about 50 billion neurons, representing about one half of all the neurons in the brain. The cerebellum is known to be highly somatotopic, but many details of the somatotopy in the human cerebellum are still unknown, and investigations into this relationship in the in vivo human model is difficult. Through the use of modern imaging methods and image analysis tools to investigate subjects with a specific set of genetically defined cerebellar ataxias, we have a unique opportunity to examine the relationships between regional atrophy and motor and cognitive dysfunction. Atrophy is a complex process, however, and its detailed study requires the characterization of both gray matter and white matter tissues in the cerebellum itself and in tightly connected structures. Furthermore, it is critical to analyze both lobular differences as well as medial-to-lateral differentiation because of the currently theorized organization of cerebellar somatotopy. These requirements preclude the use of manual delineation (which has been used in most previous studies) for anything except very small studies, due to its variability and high cost. As well, detailed analysis of white matter requires the use of diffusion-weighted magnetic resonance images (MRI), which are notoriously difficult to analyze manually.
The cerebellum touches on virtually all aspects of human experience, and its dysfunction causes a host of debilitating symptoms. New medical image computing methods to study the cerebellum in humans for both scientific and clinical purposes are needed. This workshop will provide a snapshot of the current progress in cerebellum research based on technical methods. Cerebellum lobule segmentation and labeling will be a major theme, especially given the focus of the Challenge, and opportunities for extended discussions will be provided on this and other topics related to the cerebellum. Researchers choosing to be involved in the Challenge portion will get an opportunity to characterize their cerebellum labeling method on a newly released multi-site dataset in a grand challenge format. The training datasets will be released to all researchers after the completion of this challenge.