A deep learning based convolutional neural network model with VGG16 feature extractor for the detection of Alzheimer Disease using MRI scans

Abstract

Alzheimer's disease (AD) is one of the most prevalent types of dementia, which primarily affects people over age 60. In clinical practice, it is a challenging task to identify AD in its early stages, and there are currently very few reliable diagnostic systems available for the identification. Additionally, clinical studies for AD medications have a high risk of failure, and currently, there is no confirmed cure. There are various stages of AD: very mild demented, mild, and moderate. It is challenging to identify these stages, due to which the very mild demented case worsens and results in a complete health loss along with weak memory and makes it unable to perform daily tasks without the assistance of others. Early identification of mild cases can help patients to guide additional medical care to stop the disease's progression and avoid brain damage. Recently, there has been a substantial amount of interest in applying deep learning (DL) for early AD recognition. The limitations of these algorithms are that they cannot detect changes in the brain networks in patients with mild demented functional working brain networks. However, to stop AD growth, various scientists and researchers are striving to build methods for early identification by using MRI images. In this article, for early diagnoses of AD, two MRI datasets containing 6400 and 6330 images have been used, and the DL algorithm is utilized by applying a neural network classifier with a VGG16 feature extractor for the early diagnosis of AD, which results in the outcome in the form of accuracy, precision, recall, AUC and F1-score as (90.4%, 0.905, 0.904, 0.969, and 0.904), and (71.1%, 0.71, 0.711, 0.85, and 0.71) for dataset 1 and dataset 2, respectively. Furthermore, the results are compared with previous studies, which concluded the proposed model performs better. Lastly, this article is applicable to identify various machine learning (ML) and DL approaches that can be utilized to study AD stage identification.