Geometric invariants for classification of cortical sulci

We have developed a computational method based on a family of geometric measures for the purpose of classification and identification of families of sulcal curves from human brain surfaces. Topologically correct cortical surfaces of the human brain were extracted from magnetic resonance images. Polygonal curves representing sulcal curves were then generated on each surface. Geometric measures including Gauss integrals, moments and topological features were computed for each curve to obtain a set of feature vectors in a high dimensional vector space. These feature were used to classify the curves into sulcal and hemispheric classes. In our preliminary results, an automatic differentiation between sulcal paths from the left or right hemispheres and individual sulcal curve classification were achieved, indicating these measures may have biological significance in neuroscientific data.