Enhancing Performance of Convolutional Neural Network Using Data Augmentation for Alzheimer's Disease Classification

Résumé fourni par la source

Blood samples are needed to identify Alzheimer's disease (AD)-associated genes for early diagnosis. There are no effective gene profile-based AD biomarkers. Recent research has focused on using genetic data to predict disease using AI methods like machine learning (ML) and deep learning (DL). Several diseases have received substantial attention, but AD has not, possibly due to a lack of effective methods. For automated AD classification using gene expression data, an optimum DL-based method is presented to address these difficulties. However, the main problem with using DL methods is that they require a large number of data samples for better accuracy. AD classification utilizing gene expressions is mostly hindered by insufficient gene expression data. The objective is to create a deep learning system for automatic AD categorization utilizing data augmentation. The pre-trained Convolutional Neural Network (CNN) is initially utilized for automatic feature extraction from the obtained gene expressions. The SoftMax classifier is developed for disease prediction. Automated CNN layers and feature optimization enable gene feature selection. A few fully linked layers are created before using the ML classifier. AD classification uses fully linked layers and SoftMax classifier. To enhance the overall classification performances, acquired gene expressions are augmented using six different methods. The gene expressions are augmented using erase, flip, drop, sine, cutout, and shift. The performance of the proposed model is evaluated using two research datasets.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.