European Biotechnology Congress, Valencia, Spain, 11 - 13 April 2019, vol.305
According to the Alzheimer's Association, Alzheimer’s
disease (AD) is the most common type of dementia (%60-80) that affects memory,
language, judgment, etc. [1]. Clinical diagnosis is the gold
standard for AD [2]. Mild cognitive impairment (MCI) is a phase between usual forgetfulness
due to aging and AD. Every individual with MCI does not develop AD. Magnetic Resonance
Imaging (MRI) can detect brain anomalies related to MCI and AD and might be
used to differentiate these two conditions.
In this study our goal was to perform automatic
classification of healthy, MCI and AD patients using volumetric features
extracted from segmented MRI scans. Here, we employed labeled MR images from Alzheimer's
Disease Neuroimaging Initiative (ADNI) database. There were a total of 351 MRI
scans from 22 healthy subjects, 20 MCI patients, and 18 AD patients. The slices
from MRI scans were segmented using CAT12 toolbox in Statistical Parametric
Mapping (SPM) software implemented in MATLAB. After the segmentation, SPM
provided us white matter, grey matter, cerebrospinal fluid, total intracranial
volume and cortical thickness values, which we call MRI volumetric features.
Along with sex and age information, volumetric features were employed as the
input for the classification phase of our study. The methods investigated here
were k-nearest neighbors (kNN), random forest, naïve Bayes, and support vector machines.
In the classification, kNN performed the best among other methods resulting in 91.5%
accuracy with 10-fold cross-validation. The number of correctly classified
instances was 321 out of 351 scans (71/77 AD, 134/145 MCI, 116/129 healthy).
This study showed that volumetric features have a
great potential in automatic discrimination of healthy, MCI and AD patients.