Using Workbench Command

FILTER CLUSTERS BY AREA/VOLUME
   wb_command -cifti-find-clusters
      <cifti> - the input cifti
      <surface-value-threshold> - threshold for surface data values
      <surface-minimum-area> - threshold for surface cluster area, in mm^2
      <volume-value-threshold> - threshold for volume data values
      <volume-minimum-size> - threshold for volume cluster size, in mm^3
      <direction> - which dimension to use for spatial information, ROW or
         COLUMN
      <cifti-out> - output - the output cifti

      [-less-than] - find values less than <value-threshold>, rather than
         greater

      [-left-surface] - specify the left surface to use
         <surface> - the left surface file

         [-corrected-areas] - vertex areas to use instead of computing them
            from the surface
            <area-metric> - the corrected vertex areas, as a metric

      [-right-surface] - specify the right surface to use
         <surface> - the right surface file

         [-corrected-areas] - vertex areas to use instead of computing them
            from the surface
            <area-metric> - the corrected vertex areas, as a metric

      [-cerebellum-surface] - specify the cerebellum surface to use
         <surface> - the cerebellum surface file

         [-corrected-areas] - vertex areas to use instead of computing them
            from the surface
            <area-metric> - the corrected vertex areas, as a metric

      [-cifti-roi] - search only within regions of interest
         <roi-cifti> - the regions to search within, as a cifti file

      [-merged-volume] - treat volume components as if they were a single
         component

      [-size-ratio] - ignore clusters smaller than a given fraction of the
         largest cluster in the structure
         <surface-ratio> - fraction of the structure's largest cluster area
         <volume-ratio> - fraction of the structure's largest cluster volume

      [-distance] - ignore clusters further than a given distance from the
         largest cluster in the structure
         <surface-distance> - how far from the largest cluster a cluster can
            be, edge to edge, in mm
         <volume-distance> - how far from the largest cluster a cluster can be,
            edge to edge, in mm

      [-start] - start labeling clusters from a value other than 1
         <startval> - the value to give the first cluster found

      Outputs a cifti file with nonzero integers for all brainordinates within
      a large enough cluster, and zeros elsewhere.  The integers denote cluster
      membership (by default, first cluster found will use value 1, second
      cluster 2, etc).  Cluster values are not reused across maps of the
      output, but instead keep counting up.  The input cifti file must have a
      brain models mapping on the chosen dimension, columns for .dtseries, and
      either for .dconn.  The ROI should have a brain models mapping along
      columns, exactly matching the mapping of the chosen direction in the
      input file.  Data outside the ROI is ignored.