Hybrid Deep and Machine Learning Framework for Predicting Alzheimer’s Disease

Abstract

Dementia is term related to many symptoms regarding brain abilities for old people. These symptoms include losing memory and thinking abilities. There are many causes leading to dementia, such as vascular dementia, Parkinson’s disease, and also severe head injury. But one of the biggest reasons is Alzheimer’s disease. Diagnostic of Alzheimer’s is challenging for the psychiatrists. There are many ways to diagnostic Alzheimer’s from conducting tests for memory to thinking skills to being evaluated by a healthcare professional. Brain-imaging as MRI, can be used to diagnose Alzheimer’s dementia earlier. This paper proposes a hybrid model to predict Alzheimer’s early by combining different machine learning (ML) models with deep learning models. Many models in this hybrid are used to get the powerful from each model and increasing the accuracy and to overcome the shortage of other models if it exist. We use two datasets of MRI for the brain from Kaggle. The result shows some hybrid models achieved outstanding results, as MobileNet with KNN scores the highest accuracy of 0.96, precision of 0.96, recall of 0.96, and F1-score of 0.96. This suggests that KNN is highly effective in leveraging the MobileNet. These top classifiers from the hybrid models indicate that combining robust feature extractors such as MobileNet, InceptionV3, and VGG16 with effective ML algorithms such as KNN, MLP, and random forest (RF) provides the best results for Alzheimer’s disease prediction.