High Accuracy Diagnosis for MRI Imaging Of Alzheimer’s Disease using Xgboost
Esraa M. Arabi1, *, Ashraf S. Mohra1, Khaled S. Ahmed1
Identifiers and Pagination:Year: 2022
E-location ID: e187407072208300
Publisher ID: e187407072208300
Article History:Received Date: 16/3/2022
Revision Received Date: 2/6/2022
Acceptance Date: 17/6/2022
Electronic publication date: 19/10/2022
Collection year: 2022
open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: https://creativecommons.org/licenses/by/4.0/legalcode. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Alzheimer’s disease (AD) is the most epidemic type of dementia. The cause and treatment of the disease remain unidentified. However, when the impairment is still at a preliminary stage or mild cognitive impairment (MCI), the symptoms might be more controlled, and the treatment can be more efficient. As a result, computational diagnosis of the disease based on brain medical images is crucial for early diagnosis.
In this study, an efficient computational method was introduced to classify MRI brain scans for patients with Alzheimer’s disease (AD), mild cognitive impairment (MCI), and normal aging control (NC), comprising three main steps: I) feature extraction, II) feature selection III) classification. Although most of the current approaches utilize binary classification, the proposed model can differentiate between multiple stages of Alzheimer’s disease and achieve superior results in early-stage AD diagnosis. 158 magnetic resonance images (MRI) were taken from the Alzheimer’s Disease Neuroimaging Initiative database (ADNI), which were preprocessed and normalized to be suitable for extracting the volume, cortical thickness, sulci depth, and gyrification index measures for various brain regions of interest (ROIs), as they play a considerable role in the detection of AD. One of the embedded feature selection method was used to select the most informative features for AD diagnosis. Three models were used to classify AD based on the selected features: an extreme gradient boosting (XGBoost), support vector machine (SVM), and K-nearest neighborhood (KNN).
Results and Discussion:
XGBoost showed the highest accuracy of 92.31%, precision of 0.92, recall of 0.92, F1-score of 0.92, and AUC of 0.9543. Recent research has reported using multivariable data analysis to classify dementia stages such as MCI and AD and employing machine learning to predict dementia stages.
In the proposed method, we achieved good performance for early-stage AD (MCI) detection, which is the most targeted stage to be identified. Moreover, we investigated the most reliable features for the diagnosis of AD.