3140
View Abstract / View Presentation
Group cohesive parcellation results in superior functional-based parcellation with greater parcel-level parsimony than current approaches
Ajay Nemani1 and Mark Lowe1
1Cleveland Clinic, Cleveland, OH, United States
Group cohesive parcellation is introduced, yielding parcels with exemplar time series that highly correlate to their members at group and individual level.  Group cohesive parcellation also compares favorably to existing parcellations using common measures of cluster validity.
Distribution of group parcel cohesion (mean) vs group parcel dispersion (std). Most parcels are represented as a histogram (gray scale), the largest 20 parcels are shown individually (color, size proportional to parcel size). The overall size-weighted group mean (0.58, black line) and minimum (0.5, grey line) are also shown. a) Corresponding spatial map. Distributions of cohesion shown for the current (b), anatomical (c), and connectivity-based (d) parcellations when projected to each subject’s data. GCP = group parcel cohesion, Des = Destrieux6, YC = Yeo/Choi7,8.
Parcels are validated across all 18 subjects based on parcel cohesion (top row), homogeneity (middle row), and modified silhouette (bottom row) for group cohesive (left column), anatomical, and connectivity-based (right column) parcellations. Boxes represent median and interquartile ranges, whiskers cover 95% of the distributions. Group mean are also shown (black lines and bottom right values). GCP = group parcel cohesion, Des = Destrieux6, YC = Yeo/Choi7,8.