Hippocampal segmentation using structured extreme learning machine with bag of features

Automatic hippocampal segmentation is one of the technique used by physicians to extract hippocampal in order to help them in diagnosing brain related diseases. Previous research shows that hippocampal could be segmented either by bounding box or atlas template, but both of these methods has been criticized as it depends heavily on normalization result between MRI subject and MRI template. In this paper, we introduce an automatic segmentation technique where Structured Extreme Learning Machine (S-ELM) will be used to segment hippocampal. The objective of this paper is mainly to investigate the performance of the S-ELM where every learning hyperparameters that will be used in this study will be analyzed. The proposed technique will also employ Bag of Feature (BoF) as the feature extraction method. Constructing BoF can be based on feature point location through salient point, regular grid, random point and the combination of all aforementioned feature point locations to locate the hippocampal. In order to validate the performance of the proposed framework, the investigation will be carried out using ADNI dataset that can be obtained from http://adni.loni.usc.edu/. The results show that S-ELM can locate the hippocampal region by using grid point selection method compare with another feature point that we have proposed.