Acapulco Parcellation

The convolutional neural network learns patterns from structural MRI scans that have been manually labeled by anatomical lobule. It then applies those learned patterns to T1-weighted images, identifying cerebellar boundaries and assigning an anatomical label to each voxel. This transfers expert-defined organization into an automated process that can be applied consistently across participants.

Voxel-level labeling provides a detailed map in which each location receives an anatomical assignment rather than treating the cerebellum as one undivided structure. This supports measurements and comparisons at the level of defined lobules, making individual anatomical variation easier to examine. The resulting organization is especially useful when studies compare cerebellar structure across many participants.

Manual tracing requires researchers to delineate cerebellar boundaries directly for each scan, whereas Acapulco Parcellation automates boundary detection and anatomical assignment after learning from manually labeled examples. The automated approach reduces the time and subjectivity associated with repeated tracing. Its standardized outputs can therefore make comparisons across participants more reproducible.

The method analyzes structural, T1-weighted magnetic resonance images. A convolutional neural network examines the image data, identifies boundaries within the cerebellum, and produces voxel-level labels corresponding to anatomically defined lobules. In a research workflow, these labels provide a structured representation of cerebellar organization that can be used for consistent downstream analysis.

Neuroscientists can use the technique when they need consistent cerebellar measurements across participants or large groups of scans. It is relevant to studies of normal individual anatomical variation and to investigations of structural changes associated with neurological or psychiatric conditions. Automating the labeling process helps such projects analyze cerebellar organization without relying entirely on repeated manual tracing.

Standardized labels allow researchers to organize structural findings according to the same cerebellar lobules across participants. This supports clearer comparisons between individuals and can strengthen analyses of condition-associated anatomical changes. Because the outputs are generated through a consistent automated procedure, the method is also suited to larger-scale neuroscience research focused on cerebellar structure.