Blockchain-Based Deep Learning Approach for Alzheimer's Disease Classification

Introduction

AD is a kind of dementia that often affects the elderly and is characterized by gradual cognitive impairment and a deterioration in the brain's functioning skills. According to the 2016 World Alzheimer Report, around 46.8 million individuals have AD and dementias which is illustrated as Figure 1. It is anticipated that the incidence of Alzheimer's will double every 20 years and that by 2050, the global prevalence of Alzheimer's will reach around 131.5 million (Ding et al.,2019). Approximately 60% of the brain's total volume is contained within this area. Grey matter, situated deep inside the brain, performs this crucial processing. This structure consists of the dendrites and nuclei of neurons. It accounts for around 40% of the volume of the brain. The white and grey matter of the central nervous system and spinal cord are protected from mechanical shocks by cerebrospinal fluid. Different hormones released by the hypothalamus facilitate communication between white and grey matter in the central nervous system (Zhang et al.,2019). Artificial Intelligence (AI) encompasses a vast array of algorithms and methodologies, including genetic algorithms (Wang et al.,2019; Alberdi & Weakley, 2018; Li et al.,2019), neural networks (Pavisic et al.,2021), and evolutionary algorithms (López-De-Ipiña et al., 2020; Sapey-Triomphe et al., 2015; Tzimourta, n.d.; Afrantou et al., 2019).

Figure 1.

People affected by AD according to age

The most obvious symptoms include ineffective communication, increased susceptibility to infection, poor judgment, poor sense of direction, short-term memory loss, and visual difficulties. Recent research indicates that approximately 50 million people worldwide have Alzheimer's (Krishna et al., 2019; Chandra et al., 2019; Ke et al., 2019; Zheng et al., 2019).

However, most existing DL methods train a deep convolutional neural network (CNN) model from scratch and suffer from limitations (Lian et al., 2020). First, it requires a substantial quantity of labeled training data, which may be difficult to acquire in the medical domain. Second, it requires enormous computational and memory resources; training a model would take longer without them. Thirdly, it requires careful optimization of network parameters via regularisation; failure to do so leads to overfitting or underfitting issues. It is also essential to ensure that the trained model generalizes well to unknown data (Maggipinto et al., 2017).

Using functional magnetic resonance imaging (fMRI) images to train a deep neural network with blockchain to detect a wide variety of brain events (Chitradevi & Prabha, 2020; Lu et al., 2018), we achieved this goal. Due to advancements in the scientific and technical sectors, computer-aided diagnostic (CAD) systems have been developed. These systems aid in the analysis of medical imaging by researchers and clinicians.

In the early phases of AD, computerized diagnosis is essential for human health. Because AD is a neurological disorder, its incubation period is lengthy. Consequently, it is essential to monitor AD symptoms at various times. Using a variety of image categorization techniques can improve the accuracy of AD diagnosis.

DL researchers have discovered other ways to identify the severity of AD in individuals using MRI scan images (Gorges et al., 2018; Yu et al., 2020). In terms of image processing and analysis, the higher the image quality, the more accurate the results.

The rest of this article is structured as follows: the “Related Work” section reviews pertinent works. The “Problem Statement and Solution Plan” section summarizes the study's primary concerns and objectives. The section “Materials and Methods” discusses the proposed method. The section “Experimental Outcomes and Model Evaluation” evaluates the experiments and their results. The section “Conclusion” finalize the article.