Open AccessDevelopment of a Predictive Analytic Tool for Detection of Alzheimer’s Disease
Author(s) -
Aw Hui Yee,
Azian Azamimi Abdullah,
Juhaida Abu Bakar
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1997/1/012004
Subject(s) - feature selection , random forest , machine learning , artificial intelligence , computer science , regularization (linguistics) , lasso (programming language) , disease , elastic net regularization , support vector machine , artificial neural network , boosting (machine learning) , magnetic resonance imaging , medicine , pathology , radiology , world wide web
Alzheimer’s disease (AD) is the common neurodegenerative disease that causes the impairment of the brain tissue with a gradual decline in the memory and thinking skills among the elderly population. However, there is a lack of reliable biomarker and effective medical treatment to detect and cure the AD respectively. Since there is a commitment agrees that AD should be the focus on the early diagnosis stage due to the great improvement on the treatment efficacy, hence with the help of imaging modality such as Magnetic Resonance Imaging (MRI), it can provide a detailed brain’s features information for analysis and interpretation purposes. Besides that, this project is focusing on the implementation of the supervised machine learning algorithms on the prediction of AD based on the Graphical User Interface (GUI). By referring to the results obtained from classifiers’ performance evaluation, Random Forest achieved the best performance in classification of AD with the highest accuracy of 0.7825 and AUC score of 0.8314 along with Lasso Regularization, followed by Gradient Boosting Machine (GBM) with accuracy of 0.7352 and Deep Neural Network (DNN) with accuracy of 0.6201 only. The machine learning with feature selection, Lasso Regularization is performed much better than those without feature selection. Thus, the GUI is then developed to provide convenience for the user to accurately predict their current AD’s status.