Synthetic data analysis for early detection of Alzheimer progression through machine learning algorithms.

Journal: PeerJ. Computer science
Published Date:

Abstract

Alzheimer's disease (AD) is a serious neurodegenerative disorder that causes incurable and irreversible neuronal loss and synaptic dysfunction. The progress of this disease is gradual and depending on the stage of its detection, only its progression can be treated, reducing the most aggressive symptoms and the speed of its neurodegenerative progress. This article proposes an early detection model for the diagnosis of AD by performing analyses in Alzheimer's progression patient datasets, provided by the Alzheimer's Disease Neuroimaging Initiative (ADNI), including only neuropsychological assessments and making use of feature selection techniques and machine learning models. The focus of this research is to build an ensemble machine learning model capable of early detection of a patient with Alzheimer's or a cognitive state that leads to it, based on their results in neuropsychological assessments identified as highly relevant for the detection of Alzheimer's. The proposed approach for the detection of AD is presented with the inclusion of the feature selection technique recursive feature elimination (RFE) and the Akaike Information Criterion (AIC), the ensemble model consists of logistic regression (LR), artificial neural networks (ANN), support vector machines (SVM), K-nearest neighbors (KNN) and nearest centroid (Nearcent). The datasets downloaded from ADNI were divided into 13 subsets including: cognitively normal (CN) subjective memory concern (SMC), CN early mild cognitive impairment (EMCI), CN late mild cognitive impairment (LMCI), CN AD, SMC EMCI, SMC LMCI, SMC AD, EMCI LMCI, EMCI AD, LMCI AD, MCI AD, CN AD and CN MCI. From all the feature results, a custom model was created using RFE, AIC and testing each model. This work presents a customized model for a backend platform to perform one-versus-all analysis and provide a basis for early diagnosis of Alzheimer's at its current stage.

Authors

  • Ana Gabriela Sánchez Reyna
    Systems and Computing Department, TecNM/Technological Institute of Aguascalientes, Aguascalientes, Aguascalientes, Mexico.
  • Ricardo Mendoza-Gonzalez
    Systems and Computing Department, TecNM/Technological Institute of Aguascalientes, Aguascalientes, Aguascalientes, Mexico.
  • Huizilopoztli Luna-García
    Unidad Académica de Ingeniería Eléctrica, Universidad Autónoma de Zacatecas, Zacatecas, Zacatecas, Mexico.
  • José María Celaya Padilla
    Unidad Académica de Ingeniería Eléctrica, Universidad Autónoma de Zacatecas, Zacatecas, Zacatecas, Mexico.
  • Jorge Alejandro Morgan Benita
    Unidad Académica de Ingeniería Eléctrica, Universidad Autónoma de Zacatecas, Zacatecas, Zacatecas, Mexico.
  • Carlos H Espino-Salinas
    Unidad Académica de Ingeniería Eléctrica, Universidad Autónoma de Zacatecas, Zacatecas, Zacatecas, Mexico.
  • Jorge I Galván-Tejada
    Unidad Académica de Ingeniería Eléctrica, Universidad Autónoma de Zacatecas, Zacatecas, Zacatecas, Mexico.
  • David Rondon
    Estudios Generales, Universidad Continental, Arequipa, Peru.
  • Klinge Villalba-Condori
    Vicerrectorado de Investigación, Universidad Nacional Pedro Henríquez Ureña, Santo Domingo, Dominican Republic.

Keywords

No keywords available for this article.