A Multifunctional Approach to Feature Extraction from fMRI Im- ages in Alzheimer's disease

Shahriar  Mohammadi; Soraya  Zarei; Ali Mohammad  Mosadeghrad

doi:10.18502/jbe.v11i2.20556

Shahriar Mohammadi Information Technology Group, Department of Industrial Engineering, K.N. Toosi University of Technology, Tehran, Iran.
Soraya Zarei Information Technology Group, Department of Industrial Engineering, K.N. Toosi University of Technology, Tehran, Iran.
Ali Mohammad Mosadeghrad Department of Health Management and Economics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran

DOI: https://doi.org/10.18502/jbe.v11i2.20556

Keywords: Medical imaging; Brain diseases; fMRI images; Feature extraction

Abstract

Introduction: The use of fMRI imaging in medical science has led to the diagnosis of diseases at the very first stages before the disease get advancedwhich plays a significant role in some diseases such as Alzheimer's. Extracting useful information from these images is the first step in the initial diagnosis of the disease that the accuracy in extracting as much of this information as possible contributes significantly to the initial diagnosis. Increases the speed of processing and estimation accuracy which was done in the present study using a multi-purpose method. While in recent studies, simpler methods with a limited number of features were used.

Methods: The information of 140 patients with Alzheimer's disease was obtained, and the stable multipurpose feature extraction method was used to extract the information. In this way, two-level wavelet, modeling of wavelet coefficients, normalization method and feature selection are applied.

Results: The results obtained from the examination of 285 features in five categories showed that some of the information contained in the features overlapped and lacked useful information. In addition, dimensionality and noise reduction using the PCA algorithm showed that about 41% of the relevant features are outliers or missing information.

Conclusion: In general, increasing the speed of processing and estimation accuracy which was done in the present study using a multi-purpose method. While in recent studies, simpler methods with a limited number of features were used.