%0 Journal Article
%T Alzheimer¡¯s Disease Detection in Brain Magnetic Resonance Images Using Multiscale Fractal Analysis
%A Salim Lahmiri
%A Mounir Boukadoum
%J ISRN Radiology
%D 2013
%R 10.5402/2013/627303
%X We present a new automated system for the detection of brain magnetic resonance images (MRI) affected by Alzheimer¡¯s disease (AD). The MRI is analyzed by means of multiscale analysis (MSA) to obtain its fractals at six different scales. The extracted fractals are used as features to differentiate healthy brain MRI from those of AD by a support vector machine (SVM) classifier. The result of classifying 93 brain MRIs consisting of 51 images of healthy brains and 42 of brains affected by AD, using leave-one-out cross-validation method, yielded classification accuracy, 100% sensitivity, and specificity. These results and a processing time of 5.64 seconds indicate that the proposed approach may be an efficient diagnostic aid for radiologists in the screening for AD. 1. Introduction Alzheimer¡¯s disease (AD) is a progressive and degenerative disease that affects brain cells, and its early diagnosis has been essential for appropriate intervention by health professionals. Noninvasive in vivo neuroimaging techniques such as magnetic resonance imaging (MRI) and positron emission tomography (PET) are commonly used to diagnose and monitor the progression of the disease and the effect of treatment. In this regard, the problem of developing computer aided diagnosis (CAD) tools to distinguish images with AD from those of normal brains has been extensively addressed in the past years [1¨C12]. A review of a recent work follows. Magnin et al. [1] used the relative weight of gray matter versus white matter and cerebrospinal fluid in 90 regions of interests (ROI) as features classified with SVM. Based on the bootstrap method, the SVM obtained 94.5% average classification accuracy in the classification of 16 AD and 22 control (healthy) subjects, with a mean specificity of 96.6% and a mean sensitivity of 91.5%. Ram¨ªrez et al. [2] proposed a classification system for AD based on the partial least square (PLS) regression model for feature extraction (identification of discriminative voxels) and the random forest (RF) classifier. The PLS-RF system yielded accuracy, sensitivity, and specificity values of 96.9%, 100%, and 92.7%, respectively, after classifying 41 normal and 56 AD images using the leave-one-out cross-validation method. Salas-Gonzalez et al. [3] used Welch¡¯s -test to identify voxels that provide higher difference between normal and AD images. The identified voxels formed the main features to classify, and a SVM with linear kernel reached 96.2% accuracy in distinguishing 41 normal and 38 AD images using leave-one-out cross-validation method; the sensitivity and the
%U http://www.hindawi.com/journals/isrn.radiology/2013/627303/