slugMR Brain Image Segmentation Based on Principle Component Analysis and Self-Organizing Map
Journal Title: International Journal for Research in Applied Science and Engineering Technology (IJRASET) - Year 2014, Vol 2, Issue 3
Abstract
In this paper, a fully unsupervised segmentation of Magnetic Resonance (MR) brain image is presented, which is based on a competitive learning algorithm- Self Organizing Map (SOM). We tried to address the problem of segmentation of MR brain images using Principle Co mponent Analysis (PCA) and unsupervised classifier. The proposed technique contains number of steps such as preprocessing using Brain Extraction Tool (BET), feature extraction (first and second order features), feature selection using PCA and segmentation using SOM clustering. Our proposed method is performed over real MR data provided by Internet Brain Repository (IBSR) database. Performance evaluation using Tanimoto performance index shows that the proposed method has good segmentation results. Tanimoto p erformance index gives mean and standard deviation of 0.59 0.06 for white matter (WM) and 0.58 0.05 for gray matter (GM). Algorithm offers 86.74% and 70.2% sensitivity for WM and GM respectively and 96.49% and 96.44% specificity for WM and GM. This fully unsupervised method can be used to identify the brain disorders such as brain tumours, dementia Alzheimer’s disease and other neuro anatomical disorders.
Authors and Affiliations
Jesna M, Kumudha Raimond
Keywords
Related Articles
- EP ID EP17797
- DOI -
- Views 330
- Downloads 12
Optical and Mechanical Properties of KDP - ADP Mixed Crystals
KDP and ADP single and mixed crystals have been grown by slow evaporation solution technique. The solubility of data KDP and ADP were determined and structures, optical, mechanicals properties of the crystals have been...
Comparative Analysis on Environmental and Economic Parameters of Biomass and Coal Fired Thermal Power Plant
This paper is trying to focus our attention to fulfil our current energy demands in sustainable way rather than the conventional methodology that was prevalent before independence in our country. One of the finest ways...
slugAn Efficient Priority Based Task Management In Grid Computing-2
Grid computing can integrate and utilize heterogeneous computing resources which are connected through networks without the limitation of geography.In this research paper we have proposed a priority based task schedulin...
slugEfficient Design of Power and Clock network for Large Chip Design Using Energy Recycling Technique
This project aim is to develop a low- skew in multigiga hertz clocks circuit in operation with giant digital chips. A capacitance are often charged from the availability then discharged to ground. In giant digital chi...
Optimization of Forming Process Parameters for Appropriate Distribution of Wall Thickness in Sheet Metal Component
While obtaining the final sheet metal component the defects are occurred in sheet metal forming process which are reduced by varying the forming process parameters by trial and error method. This causes loss in terms of...