Feature Selection Based on Machine Learning in MRIs for Hippocampal Segmentation

الملخص EN

Neurodegenerative diseases are frequently associated withstructural changes in the brain.

Magnetic resonance imaging (MRI)scans can show these variations and therefore can be used as a supportivefeature for a number of neurodegenerative diseases.

The hippocampushas been known to be a biomarker for Alzheimer disease and other neurologicaland psychiatric diseases.

However, it requires accurate, robust,and reproducible delineation of hippocampal structures.

Fully automaticmethods are usually the voxel based approach; for each voxel a numberof local features were calculated.

In this paper, we compared four differenttechniques for feature selection from a set of 315 features extractedfor each voxel: (i) filter method based on the Kolmogorov-Smirnov test;two wrapper methods, respectively, (ii) sequential forward selection and(iii) sequential backward elimination; and (iv) embedded method basedon the Random Forest Classifier on a set of 10 T1-weighted brain MRIsand tested on an independent set of 25 subjects.

The resulting segmentationswere compared with manual reference labelling.

By using only23 feature for each voxel (sequential backward elimination) we obtainedcomparable state-of-the-art performances with respect to the standardtool FreeSurfer.