Hippocampal shape analysis in epilepsy using Laplace-Beltrami spectrum

Shape analysis plays an important role in many medical imaging studies. One of the recent shape analysis methods uses the Laplace Beltrami eigenvalues which is also used in this paper for global shape comparison of hippocampus of normal subjects and epileptic patients. Popularity of the Laplace Beltrami operator in this field is due to its isometry-invariance which avoids pre-processing steps like mapping, registration, and alignment. In addition, it is capable of revealing fine details in shapes that makes this method a good choice for deformation detecting purposes like epilepsy diagnosis. To examine capability of the proposed method, statistical analysis and two ways of classification, support vector machine (SVM) and finding out of normal range (ONR) subjects, are used. Moreover, to evaluate our classification results, K-fold cross-validation is performed. The best achieved results were true positive rate of 91.9% and false positive rate of 33.3%, yielded by ONR classifiers using 3 selected eigenvalues.