Deep learning for early diagnosis of Alzheimer’s disease: a contribution and a brief review
| Autor: | Wellington Pinheiro dos Santos, Rodrigo Gomes de Souza, Manoel Eusebio de Lima, Washington Wagner Azevedo da Silva, Iago Richard Rodrigues Silva, Gabriela dos Santos Lucas e Silva, Ricardo Emmanuel de Souza, Roberta Andrade de Araújo Fagundes, Maíra Araújo de Santana |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
medicine.diagnostic_test
Computer science business.industry Deep learning Feature extraction Magnetic resonance imaging Cognition Machine learning computer.software_genre Convolutional neural network Random forest Support vector machine medicine Artificial intelligence business computer Classifier (UML) |
| Zdroj: | Deep Learning for Data Analytics ISBN: 9780128197646 |
| Popis: | Alzheimer’s disease (AD) is a neurodegenerative disease that results in the loss of cognitive ability of the patient. Early diagnosis of AD may contribute to a better prognosis of the disease, opening up treatment opportunities that may slow its progress. Thus, important efforts have been made to achieve early and noninvasive diagnosis. Intelligent machine learning tools have been developed to provide early diagnosis of the disease from magnetic resonance imaging (MRI) analysis. Of these tools, deep neural networks have received special attention given the success of their application in other approaches. In this chapter, we present a brief review about the diagnosis of AD based on MRI analysis using deep learning. We also present a deep architecture for early diagnosis of AD based on implicit feature extraction for the classification of MRI. This model aims to classify AD patients against a group of patients without the disease. The database used in this project is the Alzheimer’s Disease Interval Minimum Resonance (MIRIAD), to validate the proposed method. We selected 30 slices from the upper brain above the eyes for learning in this paper. The convolutional neural network (CNN) architecture is designed in three convolutional layers to extract the best features from the selected region. After that, we place the selected attributes in a vector for learning and pattern detection by another classifier in the last layer of the proposed architecture. Finally, the data is partitioned with the 10-fold cross-validation method and trained with the random forest, support vector machine (SVM), and k-nearest neighbor (k-NN) algorithms with different parameters for evaluation. The precision results are 0.8832, 0.9607, and 0.8745 for these algorithms, respectively. According to a comparative analysis performed with other state-of-the-art studies, we can highlight the efficiency and reliability of the model for early diagnosis of Alzheimer’s disease. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|