Population Classification Based on Structural Morphometry of Cortical Sulci

This paper describes a classification system discriminating male and female brains from morphometric features of cortical sulci. This system is tested on a database of 143 brains, whose sulci were automatically recognized by an artificial neuroanatomist described before. The curse of dimensionality usually plaguing classification problems is overcome using an iterative feature selection loop. The best classifier built from an optimal set of 54 morphometric features achieves a 96% correct generalization rate during a leave-one-out procedure. This result obtained using a support vector machine classifier is very appealing considering the limitations of the sulcus recognition system.

• Type: Conference
• Year: 2004
• Where: ISBI
• Authors: Edouard Duchesnay, Jean-Francois Mangin, Alexis Roche, Denis Rivière, Dimitri Papadopoulos-Orfanos, Yann Cointepas
• Keywords: Correct Generalization Rate, ISBI 2004, Iterative Feature Selection, Medical Imaging, Vector Machine Classifier