An accurate Alzheimer disease diagnosis approach based on samples balancedgenetic algorithm and extreme learning machine using MRI.

Abstract

A Samples Balanced Genetic Algorithm and Extreme Learning Machine (SBGA-ELM) designed for accurate Alzheimer Disease diagnosis and identifying biomarkers associated with AD is presented in this paper. Proposed Alzheimer Disease diagnosis approach uses set of MRI of OASIS public database to build an efficient AD classifier. Proposed Alzheimer Disease diagnosis approach contains 2 steps: 1) “voxels selection” based on Samples Balanced Genetic Algorithm (SBGA), and 2) “AD classification” based on Extreme Learning Machine (ELM). In a first step “voxels selection” step a subset of voxels with promising properties for AD diagnosis is selected from a complete set of 19879 voxels extracted from OASIS data base. A selection process is extremely complex and requires specifically designed technique. In this paper, we propose a Samples Balanced Genetic Algorithm (SBGA) for searching a subset of voxels among 19879 voxels from OASIS database. In a second “AD classification” step, a discovered subset of voxels is used to construct an efficient AD classifier based on Extreme Learning Machine (ELM). A discovered subset of voxels keeps high generalization performances of AD classification using ELM in various scenarios and highlights importance of the chosen voxels for AD research. AD classifier with maximum classification accuracy created using the best set of chosen voxels is our final AD diagnosis approach and the best set of chosen voxels is potential AD biomarkers. Experiments with proposed SBGA-ELM show an average testing accuracy 87%. Experiments clearly indicate the efficiency of the proposed SBGA-ELM for AD diagnosis and highlight improvement over existing techniques.