Dynamic Programming Applications in Defining Cortical Manifold Boundaries
J. Tilak Ratnanather
Johns Hopkins University
Sophisticated methods for constructing gray–white cortical manifolds from MRI data have emerged in recent years. The reconstructed manifolds are triangulated graphs that allow computation of principal maximal and minimal curvature at each point through fitting of the shape operator at each vertex.
We describe how dynamic programming can be used to track gyral and sulcal principal curves corresponding to the locus path between two points of maximal and minimal curvature, respectively. A third path, defined as the shortest distance between any two points irrespective of curvature, is identified as the geodesic.
These curves are used to identify and extract cortical submanifolds, enabling the construction of local coordinate systems on each submanifold. Regional cortical metrics can then be quantified using distribution maps defined with respect to these localized cortical coordinates.
An illustrative application demonstrates the use of dynamic programming to extract the planum temporale (PT) from a reconstructed superior temporal gyrus (STG) surface. The procedure tracks Heschl’s gyrus, follows the STG to the posterior ramus, computes the geodesic between relevant anatomical landmarks, and delineates the STG surface to isolate the PT region.
Presentation Files (Zip Archive)