Home > Medical Research Archives > Issue 149 > A Review of Alzheimer’s Disease and Inflammation: Pathogenesis, Inflammatory Processes, and Novel Insights from the Artificial Intelligence Perspective
![](https://esmed.org/wp-content/uploads/2022/12/MRA-cover-150x150.png)
Published in the Medical Research Archives
Jul 2023 Issue
A Review of Alzheimer’s Disease and Inflammation: Pathogenesis, Inflammatory Processes, and Novel Insights from the Artificial Intelligence Perspective
Published on Jul 29, 2023
DOI
Abstract
Alzheimer’s disease is a progressive neurodegenerative disease which is characterised by the increased deposition and spread of extracellular β-amyloid plaques and intracellular hyperphospho-rylated tau as neurofibrillary tangles. This is thought to be driven by the sustained activation of brain microglia and astrocytes. In this review, we will provide an overview of the current understanding of the pathogenesis of Alzheimer’s disease, the role of inflammation and associated factors in disease progression as well as current treatments including those in late-stage clinical trials. We will also discuss how machine learning has been previously used to create Alzheimer’s disease risk metrics and the potential for blood-based inflammatory factors to be used to create an artificial intelligence-based Alzheimer’s disease early warning system. The development of an Alzheimer’s disease-based early warning system would enable the improved use of existing and future disease-modifying agents and thereby help to slow or halt disease progression.
Author info
Abbreviations:
Aβ, Amyloid beta; AD, Alzheimer’s disease; AI, Artificial intelligence; APP, amyloid precursor protein; ARIA, Amyloid-related imaging abnormalities; ASC, Apoptosis-associated speck- like protein containing a caspase-recruitment domain; BBB, Blood-brain barrier; CSF, Cerebrospinal fluid; CNN, Convolutional neural network; CNS, Central nervous system; DL, Deep learning; FDA, Food and drug administration; IgG, Immunoglobulin; IL, Interleukin; MCI, Mild cognitive impairment; ML, Machine learning; MRI, Magnetic resonance imaging; NADPH, Nicotinamide adenine dinucleotide phosphate; NFTs, Neurofibrillary tangles; NLRC, NLR family CARD domain-containing protein; NLRP, Nucleotide-binding oligomerization domain like receptor protein; NMDA, N-methyl-D- aspartate; PET, Positron emission tomography; ROS, reactive oxygen species; SIRT, Sirtuin; sPDGFRβ, Soluble platelet-derived growth factor receptor-β; τ; Tau TLRs, Toll-like receptors.
Introduction
The objective of this review is to provide an overview of problems associated with the diagnosis of Alzheimers disease (AD), how it progresses and the effectiveness of current treatment options based on disease stage. It also explores the role of inflammation in the development and progression of AD. Additionally, we present an overview of how machine learning (ML) research has already been used to develop AD risk metrics and discuss the potential for inflammatory biomarkers to be used to create an artificial intelligence (AI) early warning system as it is important to be able to both detect and predict the onset of AD.
Alzheimers disease, the most common form of dementia, accounts for 60-80% of all cases.1 According to recent estimates, more than 6.5 million individuals are living with AD in the US alone, a number that is expected to rise to 13.8 million by 2060 due to the increasing ageing population.1 Alzheimer’s disease commonly manifests as memory impairment but may also affect expressive speech, visuospatial processing and executive functions impacting activities of daily living. Disease progression varies but on average individuals with AD live 4-8 years following their initial diagnosis.1 Alzheimers disease places an enormous burden on individuals, their families, and society in general, with estimated costs of $321 billion annually.1 Without urgent action to reduce the burden of AD these costs are set to further increase.
A key challenge for clinicians is the lack of effective treatment options for AD. Of the available treatments for AD, few affect the underlying pathology and therefore are capable of reversing disease progression. By identifying those at risk and intervening early it may be possible to halt or slow the progression of AD through a combination of pharmacological drug treatments and lifestyle changes. Initiation of treatment earlier in the disease course could also help to lower overall health care costs with potential savings of approximately $7 trillion in medical and long-term care costs.1
The identification of biomarkers that are reflective of the underlying disease pathology could aid in the early detection of AD. Already positron emission tomography (PET) and cerebrospinal fluid (CSF) concentrations of pathological proteins such as amyloid beta (Aβ) and tau (τ) are being utilised clinically in the diagnosis of AD.2 Magnetic resonance imaging (MRI) can also aid in diagnosis of AD through assessment of brain atrophy which is particularly evident in the medial temporal lobe and temporoparietal cortices.3 The routine use of these tests in clinical practice is however limited by the invasive nature of the collection procedures and the expense associated with imaging.
Other biomarkers that have been the focus of research include blood based inflammatory factors.4 Inflammation has long been known to play a key role in the pathogenesis of several neurodegenerative diseases including AD.5,6 Indeed, several inflammatory biomarkers have already been shown to be altered early in the pathogenesis of AD including soluble tumour necrosis factor receptor 2, IL-6, monocyte chemoattractant protein-1 and IL-8.7 It is hoped these biomarkers could provide a simple, non- invasive, and more cost-effective alternative to PET imaging and CSF biomarkers.
Progression of Alzheimers disease
The clinical course of AD is complex and often varies considerably between individuals. Pathologically AD is characterised by the extracellular accumulation of β-amyloid plaques and the intracellular accumulation of τ-containing neurofibrillary tangles (NFTs). Amyloid beta is derived by the proteolytic cleavage of amyloid precursor protein (APP) by γ-secretases and β- secretases. The Aβ42 peptide has a high propensity to aggregate and form higher order oligomers, fibrils, and plaques owing to the increased hydrophobicity of its C-terminus. Amyloid beta oligomers may play a direct role in AD pathogenesis by impairing both synaptic function (e.g., long-term potentiation) and synaptic structure (e.g., dendritic spines) through a range of different mechanisms.8 However, there is minimal correlation between phases of amyloid deposition and degree of cognitive decline and therefore it is unlikely that Aβ pathology alone is responsible for inducing cognitive impairment.9 In addition, evidence suggests that racial and ethnic differences found in amyloid PET positivity in people with mild cognitive impairment (MCI) and dementia may indicate differences in the underlying etiology of cognitive impairment.10 For example, the proportion of amyloid positive PET scans was greater in white than in black or Asian participants in this study. 10
According to the amyloid cascade hypothesis, deposition of Aβ may be the initiating step in AD pathology.11 Amyloid beta aggregation is thought to activate kinases which lead to the hyperphosphorylation of microtubule-associated τ protein and its polymerisation into NFTs.11 Neurofibrillary tangles typically contain a mixture of the 3R and 4R isoforms of the τ protein. Unlike Aβ pathology, τ pathology has been shown to correlate well with the degree of cognitive impairment in AD sufferers.9,12 However, the co- occurrence of Aβ, τ and abnormal microglia are a strong predictor of cognitive impairment. 13 The formation of Aβ plaques and NFTs may also contribute to AD pathology through the activation of microglial cells and astrocytes resulting in the production of inflammatory mediators. Further details about the role of inflammation in the pathogenesis of AD is discussed in the section on inflammation.14 Alzheimers disease related neuroinflammation also contributes to the neuronal loss and synaptic dysfunction seen in AD patients.
Alzheimers disease severity is generally based on the degree of cognitive impairment and ranges from mild to severe dementia. At an early stage of the disease Aβ-containing plaques can be found widely distributed throughout the isocortex while τ- NFTs are mainly observed in the allocortex of the medial temporal lobe.15 Early AD is characterised predominantly by impairments in memory which are accompanied by behavioural changes such as increased anxiety or agitation but these are often not severe enough to interfere with daily life. Alzheimers disease is particularly difficult to diagnose based on clinical symptoms alone during the early stages of the disease. As a result, CSF and PET imaging markers are commonly used along with cognitive assessments to diagnose AD. Cerebral spinal fluid biomarkers such as aggregated Aβ, total-τ, and phospho-τ, reflect key elements of AD pathophysiology and show high diagnostic accuracy even in the early stages of the disease.
As AD progresses cognitive impairment becomes more profound. Alzheimer’s disease progression is associated with the spread of τ-NFTs from the medial temporal lobe to the lateral temporal, parietal and frontal isocortex.15 At this later stage there are more obvious impairments in language, spatial cognition, executive function, and memory which do impact daily life. There are also noticeable behavioural changes such as confusion, disorientation, aggression, agitation, and at a late- stage delusions and hallucinations. Even early AD represents an advanced stage in the pathogenesis of the disease. By the time some cognitive impairment is evident significant and irreversible synaptic and neuronal loss has likely already occurred.16 The clinical phase of AD is preceded by a protracted preclinical phase in which AD pathology has been shown to accumulate 15–20 years prior to the onset of clinical symptoms.17 This preclinical stage of the disease presents a unique opportunity to identify those at risk, to start pre-emptive interventions and alter disease trajectory.
Alzheimers disease drug development pipeline There are currently seven Food and Drug Administration (FDA) approved drugs available for the treatment of AD (Table 1). These include the cholinesterase inhibitors donepezil, rivastigmine and galantamine, the N-methyl-D-aspartate (NMDA) antagonist memantine and the amyloid beta antibodies aducanumab and lecanemab. Except for the recently approved Aβ antibodies, these drugs all treat AD symptoms with little effect on the underlying disease pathology or AD progression. Even the available Aβ therapeutics have shown limited efficacy in modifying the clinical course of AD.18 However this failure may not be surprising given that Aβ plaque accumulation is already usually associated with significant neuronal changes.18
There are also numerous promising therapeutics currently in the late stages of clinical development as shown in Table 2 & 3, but these will not be discussed in detail in this review. These include other disease-modifying therapeutics targeting τ aggregates and apolipoprotein E, which could prove more effective.18 Given its role in the AD pathogenesis, inflammation is also attracting interest as a potential target in the treatment of AD. Those inflammatory targets of interest include tumour necrosis factor α, triggering receptor expressed on myeloid cells 2, CD33, proinflammatory cytokines and cyclooxygenase.19 As more becomes known about the mechanisms involved in the pathogenesis of AD undoubtedly additional targets and new therapeutic strategies will emerge.
Table 1. Food and Drug Administration approved drugs for the treatment of Alzheimers disease
Table 2. Promising small molecule therapeutics in late-stage clinical trials for Alzheimers disease
Table 3. Promising antibody therapeutics in late-stage clinical trials for Alzheimers disease
Donepezil
Donepezil is a reversible inhibitor of acetylcholinesterase which helps to increase the availability of acetylcholine at the synapses. Donepezil is used to treat symptoms associated with mild to moderate AD. A recent meta-analysis found evidence of a small benefit in cognitive function, activities of daily living and clinician‐rated global clinical state in those taking donepezil compared to placebo.20 However, they found no difference between those taking donepezil and placebo in terms of behavioural symptoms, quality of life or patient total healthcare resource utilisation.20 Higher doses (≥10mg) of donepezil were associated with a slight improvement in cognitive function but lower quality of life and more adverse events.20
Rivastigmine
Rivastigmine is a pseudo-irreversible inhibitor of acetylcholinesterase and butyrylcholinesterase, which helps to increase acetylcholine availability at the synapses. Rivastigmine is currently approved by the FDA for the treatment of symptoms associated with mild to moderate AD. A meta-analysis of studies evaluating rivastigmine compared to placebo reported better cognitive function, activities of daily living and clinicians rated global impression of changes but no change in behavioural symptoms or impact on carers.21 Additionally, patients taking rivastigmine were twice as likely to experience adverse gastrointestinal events compared to placebo, although these risks were slightly less for patients using patches compared with capsules.21
Galantamine
Galantamine is a reversible inhibitor of acetylcholinesterase, which helps to increase the availability of acetylcholine at the synapses. Galantamine is also an allosteric modulator of nicotinic cholinergic receptor sites, helping to potentiate cholinergic nicotinic neurotransmission. Galantamine is FDA-approved for the treatment of symptoms relating to mild to moderately severe AD. One meta-analysis reported a significantly greater reduction in cognition at all dosing levels in patients with AD and MCI taking galantamine compared to placebo, with the greatest effect seen over six months.22 Galantamine has similar adverse effects to the other cholinesterase inhibitors which include mainly gastrointestinal symptoms.22
Memantine
Memantine is a low affinity uncompetitive NMDA receptor antagonist. Memantine is thought to inhibit glutamate mediated neurotoxicity that occurs in response to neuronal cell death.23,24 A recent meta- analysis confirmed the efficacy of memantine with a small beneficial effect on cognition, activities of daily living and behavioural symptoms in those with moderate to severe AD.25 No clinical benefit was observed in those with mild AD and consequently it is not recommended that memantine be prescribed in these patients.25 Memantine is generally well tolerated in those with moderate to severe AD, but is associated with dizziness and headache in some patients.25
Aducanumab
Aducanumab is a human immunoglobulin (IgG) 1 antibody that binds to Aβ aggregates aiding in the removal of Aβ plaques. Aducanumab is the first drug approved by the FDA that targets the underlying pathology of AD. The FDA approval of aducanumab was initially denied due to weak efficacy data pending further confirmatory studies. Controversially, the FDA then approved aducanumab under their accelerated pathway following reanalysis of phase III data which demonstrated its ability to slow cognitive decline and reduce brain amyloid and CSF phospho-τ in patients with mild AD.26 The most common adverse event associated with aducanumab is amyloid- related imaging abnormalities (ARIA) which can indicate brain swelling.26 The phase IV trial required by the FDA recently began to confirm the clinical efficacy of aducanumab.
Lecanemab
Lecanemab is a human IgG1 antibody that binds specifically to Aβ protofibrils aiding in the removal of Aβ plaques. A phase III clinical study of lecanemab reported a reduction in markers of amyloid in early AD and moderately less decline on measures of cognition and function than placebo at 18 months.27 The most common adverse events during this study were infusion reactions and ARIA.27 As a result of its Phase II clinical data lecanemab was approved by the FDA through their accelerated approval pathway, based on the likelihood of a clinical benefit in early AD.
Neuroinflammation and Alzheimers disease
Neuroinflammation has been linked to numerous neurological and somatic disorders, such as AD, Parkinsons disease, and depression. Neuroinflammation is the immune response mechanism of the central nervous system (CNS) to various stimuli and threats, including infection, injury, and toxic protein accumulation. While an acute inflammatory response can be beneficial for clearing pathogens and cellular debris, chronic inflammation can lead to neuronal damage, synaptic dysfunction, and cognitive decline.28 Increasing evidence from genetic, epidemiological, and experimental studies indicate a strong association between neuroinflammation and AD.29– 31 Genome-wide association studies have identified several risk genes associated with AD, including ATP8B4 and ABCA1 together with rare variants of ADAM 10 gene, many of which are involved in immune processes and inflammation regulation.32–34 Furthermore, epidemiological studies suggest that long-term use of nonsteroidal anti-inflammatory drugs may reduce the risk of AD.35
Inflammation in AD occurs as a result of the binding of Aβ and activation of microglia and astrocytes receptors. Amyloid beta binding results in the production of a range of inflammatory mediators which contribute to progressive neurodegeneration as observed in AD. 36–40 Amyloid beta can also be cleared by microglia through receptor-mediated phagocytosis and degradation. The receptor for advanced glycation end product activates microglia and is involved in the production of proinflammatory molecules following Aβ binding. Activation of toll-like receptors (TLRs), particularly TLR2 and TLR4, and CD14, a member of the scavenger receptor class B family, can also do this.
However, it should be noted that the functions of microglia in the CNS are complex since in addition to their neurotoxic effects microglia can be neuroprotective.37 It is this balance between the pro- and anti-inflammatory roles of microglia which can determine the onset and progression of AD. It is persistent overactivation that can result in a proinflammatory state. A leaky blood-brain barrier (BBB) is known to contribute to the onset of AD.39,41 A leaky BBB allows the entry of fibrinogen into the brain parenchyma and its interaction with microglia and astrocytes. Fibrinogen can then interact with Aβ to form deposits in the brain parenchyma which increase the reactive microglial proinflammatory response and induce neuronal damage in AD individuals. Consequently, a leaky BBB is an early sign of cognitive dysfunction in AD sufferers.42 Using a novel cerebrospinal fluid biomarker of the BBB associated with capillary pericytes and vascular smooth muscle cells, soluble platelet-derived growth factor receptor-β (sPDGFRβ) one study found that individuals with early cognitive dysfunction develop brain capillary damage and BBB breakdown in the hippocampus irrespective of changes in Aβ and/or τ. This suggests that BBB breakdown is an early biomarker of human cognitive dysfunction, independent of Aβ or τ. In individuals with MCI increased CSF sPDGFRβ correlates with dynamic contrast-enhanced MRI measures of BBB dysfunction in specific brain regions.43
Additionally, Nation et al.42 found that brain vasculature is an important new biomarker of cognitive dysfunction in both individuals with or without Aβ and τ. There is also evidence for age- dependent BBB breakdown in the hippocampus.43 In this study, it was shown that BBB breakdown in the hippocampus and its CA1 area and dentate gyrus deteriorated with MCI that correlated with injury to BBB pericytes.42 These results enforce the idea that BBB breakdown is an early event in the ageing human brain which begins in the hippocampus, and may contribute to cognitive impairment.
Inflammatory factors and their role in Alzheimers disease
Activation of microglia and astrocytes results in the production of a range of different inflammatory mediators including cytokines, chemokines and reactive oxygen species (ROS) all of which play a role in the pathogenesis of AD. Inflammasomes cleave precursors of cytokines, such as, interleukin (IL) -1β and IL-18 to produce their active forms. An inflammasome is an intracellular multiprotein complex which is the platform for proIL-1β and proIL-18.38 Well-characterised inflammasomes are Nucleotide-binding Oligomerization Domain (NOD)-like Receptor Protein (NLRP) 1, NLRP2, NLRP3 and CARD domain-containing protein (NLRC) 4.44,45 Other components of the inflammasome include the adaptor protein, ASC (apoptosis- associated speck-like protein containing a caspase recruitment domain), which recruits the effector protein procaspase-1.38 Following recruitment of procaspase-1, it self-cleaves to form caspase-1 and cleaves the precursor of the proinflammatory molecules to form IL-1β and IL-18.38 The role of microglia and NLRP3 in the innate immune response to AD is reviewed by Hanslik and Ulland.40
Figure 1. The role of inflammation in the pathogenesis of Alzheimers disease
Both IL-1β and IL-18 are released from microglia and astrocytes and have been implicated in the onset and development of AD and the literature has been reviewed by Liu and Chan. 38,46–49 Following their release IL-1β and IL-18 can activate more astrocytes and microglia to secrete more inflammatory molecules, including cytokines, growth factors, complement molecules, chemokines and cell adhesion molecules resulting in further neuronal death. This activation may result from a microglial reaction associated with deposition of Aβ. In contrast, other cytokines, for example, IL-1 receptor antagonist, IL-4, IL-10 and transforming growth factor-β, can suppress proinflammatory cytokine production and their actions.50,51 For example, when mice showing both Aβ and τ pathology were treated with an IL-1 receptor blocking antibody, this treatment regulated brain inflammatory responses through reduction of nuclear factor kappa-light- chain-enhancer of activated B cells activity.52 There was also a partial reduction in fibrillar and oligomeric Aβ species although the overall Aβ plaque burden was unaffected.52 However, neuronal τ pathology was reduced and blocking IL- 1 signalling was shown to rescue hippocampal- dependent cognitive impairment.52
Interleukin-6 is another cytokine with a key role in brain inflammation. It can cause microgliosis and astrogliosis, and has a role in BBB integrity.53 Interleukin-6 levels are raised in AD patients and in animal models of AD but levels also rise with age. Toll like receptor 2 can bind Aβ which can then induce IL-6 production in both microglia and astrocytes and lead to the production of inflammatory mediators in astrocytes. In their review, Spooren et al.53 conclude that IL-6 cannot be labelled “beneficial” or “detrimental” in normal brain function.
Fatty acids are involved in the activation of the inflammasome and development of AD.38 For example, palmitate will activate NLRC4 in astrocytes to release IL-1β and ASC assists in the activation of the NLRC4 inflammasome. If levels of either NLRC4 or ASC are reduced in palmitate- treated astrocytes, the production of IL-1β is reduced. Both the levels of NLRC4 and ASC are upregulated in brains of individuals suffering from AD. Fatty acids are involved in the release of IL-1β from microglia.
Oxidative stress is closely linked with inflammation and contributes to many diseases including AD. Oxidative stress may play a role linking the entry of fibrinogen through a leaky BBB and subsequent synaptic and neuronal damage in AD.39 It is suggested that microglia are activated following fibrinogen binding to CD11b receptor which in turn activates nicotinamide adenine dinucleotide phosphate (NADPH) oxidase and release of neurotoxic ROS.54 The monoclonal antibody 5B8 has been found to suppress fibrin-induced NADPH oxidase, ROS and proinflammatory gene expression.54 Using a mouse animal model of AD, 5B8 entered the brain, bound to parenchymal fibrin and reduced innate immune activation and neurodegeneration.54 These authors concluded that fibrin acts as a dual inflammatory and oxidative stress signal in the brain and that fibrin immunotherapy could prevent neurodegeneration following BBB disruption.54
Neuroinflammation and mitochondrial impairment may also trigger a cycle ending in neuronal death.55 Abnormalities of mitochondria have been found in neurones and astrocytes in brains of AD patients. Both APP and Aβ accumulate in mitochondria membranes resulting in functional and structural damage.56 This may increase the number of defective mitochondria and decrease mitochondrial trafficking, resulting in abnormal mitochondrial dynamics in neurons of AD patients. There is also evidence for abnormal intracellular calcium regulation in neurons from AD patients. Artificial intelligence/machine learning approaches and inflammatory factors in Alzheimer’s disease.
With an ever-growing amount of data being collected daily, AI and ML are becoming ubiquitous in a variety of settings in modern society, healthcare being one of them.57 The processing/analysis of such a vast amount of information and the intricacies that lay behind it, are difficult tasks for single individuals and are much more suited to being handled by machines and algorithms. Tremendous progress has been made in recent years in the development and application of AI and ML tools, and these advances present an exciting opportunity for the healthcare sector. There are however several challenges that lay between state-of-the-art computational and data analytic techniques and the translation of these aspects into effective clinical practice.58 Medical data are highly sensitive private information and accessibility is often an object of debate as there are privacy concerns surrounding handling of these types of data.
Information is often extremely fragmented, being collected in different contexts by different people with different platforms and following different standards, leading to the scarcity of very high-quality large datasets. Bias is also a challenge and needs to be properly addressed to evaluate the applicability of the results obtained. In addition, lack of transparency in the interpretation of the models (especially with deep learning (DL) approaches, typically referred to as “black boxes”) often exacerbates another existing issue which is the lack of understanding from the medical professional/clinical side of what AI and ML models can and cannot do. Finally, the absence of regulations on how to use AI in healthcare complicates matters even further and in conjunction with all the above hampers the efficacious use of these approaches in clinical practice.
Nevertheless, the potential of AI in medicine has still to be fully explored and a growing number of examples are demonstrating its possibilities.59 Benefits include improvement in diagnosis and treatment of medical conditions, virtual assistance, and the more efficient use of resources by speeding up processes, reducing costs, prioritising tasks, and focusing efforts (as in the case of computational drug discovery and development), just to name a few. Due to the availability of larger and more high- quality data, ML application in AD and, in general, neurodegenerative diseases-related research has imploded in the last decade. Machine learning has the potential to play a significant role in the early prediction of neurodegenerative diseases, as it can aid in the analysis of large amounts of data, including medical imaging, genetics, and lifestyle information, to identify patterns and correlations that can lead to early detection of the disease.
A wide variety of ML approaches have been used to both understand the complex relationships between inflammatory factors and AD, as well as for developing multi-source risk metrics for such a condition. Next, we will summarise some of the interesting results in this field in terms of the different areas of ML research:
1. Association rule mining: In the context of AD and inflammation, association rule mining can help identify relationships between different inflammatory factors and AD, such as presented in the work of Chaves et al., Happwana et al. and Szalkai et al..60–62 Among other results, in this line of research Szalkai et al.62 determined some novel and also some already well-established relations connected to good or bad cognition. Among other findings, new associations of mean corpuscular haemoglobin, alkaline phosphatase and aspartate aminotransferase levels to cognition are proposed. Furthermore, it has been suggested that high cholesterol levels seem to be beneficial in an elderly population to contravene bad cognition. This type of technique has also been extrapolated to image inputs, for instance Chaves et al.60 proposed a methodology based on association rules to extract relations among activated brain areas in single photon emission computed tomography imaging. The results of this work have been used to characterise the perfusion patterns of normal subjects and to make use of them for the early diagnosis of AD (reported 94.87% classification accuracy).60
2. Supervised learning: This approach involves training a model on a labelled dataset (i.e., a dataset where the outcome of interest is known) to predict the outcome of interest for new, unlabelled data. In the context of AD and inflammation, supervised learning can be used to predict the risk of developing AD based on multiple inflammatory factors and other clinical and genetic variables.63 For instance, Kavitha et al.64 developed a variety of supervised learning models, namely decision trees, random forests, support vector machines, gradient boosting, and voting classifiers for early AD detection via longitudinal MRI data and achieved an average accuracy of 83%. Furthermore, Gaetani et al.65 used different statistical approaches and Lasso models to both identify protein biomarkers reflecting neuroinflammation in AD and determine patients with MCI due to AD (Area under the curve=0.906). Their findings suggest an association between AD pathology and the sirtuin (SIRT) 2 pathway, astrocyte and microglia activation, and BBB dysfunction, as reflected by concentration changes in the CSF of SIRT2, hepatocyte growth factor, matrix metalloproteinase 10, and CXC ligand 5.65
In the last few years, a large amount of DL-based models have been published for early detection of AD via image inputs.66,67 Cheung et al.66 developed a convolutional neural network (CNN)-based model with an 83.6% accuracy for detecting AD-dementia using retinal photographs. Furthermore, such a model was able to differentiate between participants who were Aβ positive and those who were Aβ negative - accuracies ranged from 80.6 to 89.3%.66 On the other hand, Venugopalan et al. 67 used a variety of supervised learning approaches, such as 3D CNNs, to analyse imaging (MRI), genetic (single nucleotide polymorphisms), and clinical test data to classify patients into AD, MCI, and controls. This study identified hippocampus, amygdala brain areas, and the Rey Auditory Verbal Learning Test as top distinguished features.67 Pan et al.68 designed a novel approach combining CNN architectures and Ensemble Learning and applied it to some of the most commonly acquired anatomical MRI of the brain, i.e., T1WI. The proposed model achieved average classification accuracies (± standard deviation) of 0.84 ± 0.05 for AD vs. healthy cognition and 0.79 ± 0.04 for MCI patients who will convert to AD vs. healthy cognition. A comprehensive review containing the most relevant research work done in this area (DL techniques for AD prediction) has been carried out by Shastry et al.69
3. Unsupervised learning: This approach involves identifying patterns in a dataset without prior knowledge of the outcome of interest. In the context of AD and inflammation, unsupervised learning can be used to identify subgroups of patients based on their profiles.70 In this sense, Liu et al.71 developed an unsupervised learning methodology to discriminate between scans from cognitively normal and people with AD using a limited number of labelled structural MRI scans. They used two-sample t-tests to detect the AD- relevant regions and then employed an unsupervised learning neural network to extract features from the regions. Finally, a clustering algorithm was implemented to discriminate between control and AD data based on the extracted features. The proposed method yielded an accuracy of 84%. More importantly, there is a vast amount of literature aiming to stratify AD patients into multiple subtypes via a plethora of clustering algorithms and metrics. A more detailed review of this research area is available in Alashwal et al..72 Such stratification has allowed researchers to determine several features to explain, among other things, why some individuals may progress to AD while others do not. These features include low CSF Aβ42 and impaired immediate recall.
4. Network analysis: This approach involves constructing networks of interacting variables (e.g., genes, proteins, pathways) and analysing their properties. For AD and inflammation, network analysis can be used to identify key inflammatory pathways involved in AD and to predict novel targets for drug development. For instance, Wang et al.73 developed a transcriptomic network analysis of 19 brain regions. This approach provides a comprehensive assessment of the critical molecular pathways associated with AD pathology and offers new insights into molecular mechanisms underlying selective regional vulnerability to AD at different stages of the progression of cognitive impairment.73 Analogously, Canario et al.74 utilised functional near-infrared spectroscopy to analyse brain activity in different regions and then construct minimum spanning tree-based regions to characterise the brain topologies of participants with AD, MCI, and normal controls. Their results showed the AD group to have strong correlations between the Hamilton depression rating scale and different graph metrics, suggesting a link between network organisation and the recurrence of depression in AD.
Overall, ML/AI can significantly help with the early prediction of AD by analysing large amounts of data and identifying early signs of the condition before symptoms are apparent. To achieve this goal recent research has been focused on the analysis of a large variety of multi-source databases that range from medical imaging data, such as MRI and PET scans, to cognitive tests and genetic inputs. Depending on the nature of the database, different types of algorithms have been proposed in the areas of association rule mining, supervised learning, unsupervised learning and network analysis. The application of AI/ML could also aid in the identification of inflammatory blood biomarkers which could form the basis of AD risk metrics.
Conclusion
Undoubtedly the successful treatment of AD will likely depend on the early identification and the administration of disease-modifying therapeutics before symptom onset in order to slow disease progression. Considering the clear role inflammation plays in the pathogenesis of AD, AI/ML could play an important role in helping to identify potential inflammatory blood biomarkers that could form the basis of an AD risk metric. This metric could also provide an easy method for monitoring disease progression and the effectiveness of disease-modifying therapeutics as well as enabling more tailored and personalised treatment. Such a metric could transform the way AD is managed and help relieve the significant personal and financial burden that AD places on both patients and healthcare providers.
Conflicts of interest statement
The authors have no conflicts of interest to declare.
References
1. 2022 Alzheimer’s disease facts and figures. Alzheimer’s and Dementia. 2022;18(4):700- 789. doi:10.1002/alz.12638
2. Khoury R, Ghossoub E. Diagnostic biomarkers of Alzheimer’s disease: A state- of-the-art review. Biomark Neuropsychiatry. 2019;1:100005. doi:https://doi.org/10.1016/j.bionps.201 9.100005
3. Johnson KA, Fox NC, Sperling RA, Klunk WE. Brain Imaging in Alzheimer Disease. Cold Spring Harb Perspect Med. 2012;2(4):a006213-a006213. doi:10.1101/cshperspect.a006213
4. Klyucherev TO, Olszewski P, Shalimova AA, et al. Advances in the development of new biomarkers for Alzheimer’s disease. Transl Neurodegener. 2022;11(1):25. doi:10.1186/s40035-022-00296-z
5.Guzman-Martinez L, Maccioni RB, Andrade V, Navarrete LP, Pastor MG, Ramos-Escobar N. Neuroinflammation as a Common Feature of Neurodegenerative Disorders. Front Pharmacol. 2019;10. doi:10.3389/fphar.2019.01008
6. Kinney JW, Bemiller SM, Murtishaw AS, Leisgang AM, Salazar AM, Lamb BT. Inflammation as a central mechanism in Alzheimer’s disease. Alzheimers Dement (N Y). 2018;4:575-590.
doi:10.1016/j.trci.2018.06.014
7. Shen XN, Niu LD, Wang YJ, et al. Inflammatory markers in Alzheimer’s disease and mild cognitive impairment: a meta- analysis and systematic review of 170 studies. Journal of Neurology, Neurosurgery & Psychiatry. 2019;90(5):590. doi:10.1136/jnnp-2018-319148
8. Shankar GM, Li S, Mehta TH, et al. Amyloid- β protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nat Med. 2008;14(8):837- 842. doi:10.1038/nm1782
9. Nelson PT, Alafuzoff I, Bigio EH, et al. Correlation of Alzheimer disease neuropathologic changes with cognitive status: a review of the literature. J Neuropathol Exp Neurol. 2012;71(5):362- 381. doi:10.1097/NEN.0b013e31825018f7
10. Wilkins CH, Windon CC, Dilworth-Anderson P, et al. Racial and Ethnic Differences in Amyloid PET Positivity in Individuals With Mild Cognitive Impairment or Dementia JAMA Neurol. 2022;79(11):1139. doi:10.1001/jamaneurol.2022.3157
11. Hardy JA, Higgins GA. Alzheimer’s Disease: The Amyloid Cascade Hypothesis. Science (1979). 1992;256(5054):184-185. doi:10.1126/science.1566067
12. Giannacopoulos P. Tangles and neuron numbers but not amyloid load predict cognitive status in Alzheimer’s disease. Ann Gen Hosp Psychiatry. 2003;2(1):S18. doi:10.1186/1475-2832-2-S1-S18
13. Pascoal TA, Benedet AL, Ashton NJ, et al. Microglial activation and tau propagate jointly across Braak stages. Nat Med. 2021;27(9):1592-1599. doi:10.1038/s41591-021-01456-w
14. Uddin MdS, Kabir MdT, Mamun A Al, et al. Pharmacological approaches to mitigate neuroinflammation in Alzheimer’s disease. Int Immunopharmacol. 2020;84:106479. doi:https://doi.org/10.1016/j.intimp.2020
.106479
15. Serrano-Pozo A, Frosch MP, Masliah E, Hyman BT. Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med. 2011;1(1):a006189. doi:10.1101/cshperspect.a006189
16. Gómez-Isla T, Price JL, McKeel Jr DW, Morris JC, Growdon JH, Hyman BT. Profound loss of layer II entorhinal cortex neurons occurs in very mild Alzheimer’s disease. Journal of Neuroscience. 1996;16(14):4491-4500.
17. Vermunt L, Sikkes SAM, Van Den Hout A, et al. Duration of preclinical, prodromal, and dementia stages of Alzheimer’s disease in relation to age, sex, and APOE genotype. Alzheimer’s & Dementia. 2019;15(7):888- 898.
18. Weglinski C, Jeans A. Amyloid-β in Alzheimer’s disease – front and centre after all? Neuronal Signal. 2023;7(1). doi:10.1042/ns20220086
19. Liu P, Wang Y, Sun Y, Peng G. Neuroinflammation as a Potential Therapeutic Target in Alzheimer’s Disease. Clin Interv Aging. 2022;17:665-674. doi:10.2147/CIA.S357558
20. Birks JS, Harvey RJ. Donepezil for dementia dueto Alzheimer’s disease. Cochrane Database of Systematic Reviews. 2018;2018(6). doi:10.1002/14651858.CD001190.pub3
21. Birks JS, Chong LY, Grimley Evans J. Rivastigmine for Alzheimers disease. Cochrane Database of Systematic Reviews. 2015;(9). doi:10.1002/14651858.CD001191.pub4
22. Loy C, Schneider L. Galantamine for Alzheimer’s disease and mild cognitive impairment. Cochrane Database of Systematic Reviews. 2006;(1). doi:10.1002/14651858.CD001747.pub3
23. Greenamyre JT, Maragos WF, Albin RL, Penney JB, Young AB. Glutamate transmission and toxicity in Alzheimer’s disease. Prog Neuropsychopharmacol Biol Psychiatry. 1988;12(4):421-430. doi:10.1016/0278-5846(88)90102-9
24. Livingston G, Sommerlad A, Orgeta V, et al. Dementia prevention, intervention, and care. The Lancet. 2017;390(10113):2673-2734. doi:https://doi.org/10.1016/S0140- 6736(17)31363-6
25. McShane R, Westby MJ, Roberts E, et al. Memantine for dementia. Cochrane Database of Systematic Reviews. Published online March 20, 2019.
doi:10.1002/14651858.CD003154.pub6
26. Budd Haeberlein S, Aisen PS, Barkhof F, et al. Two Randomized Phase 3 Studies of Aducanumab in Early Alzheimer’s Disease. J Prev Alzheimers Dis. 2022;9(2):197-210. doi:10.14283/jpad.2022.30
27. van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in Early Alzheimer’s Disease. New England Journal of Medicine. 2022;388(1):9-21 doi:10.1056/NEJMoa2212948
28. Ransohoff RM. How neuroinflammation contributes to neurodegeneration. Science (1979). 2016;353(6301):777-783. doi:10.1126/science.aag2590
29. Krabbe G, Halle A, Matyash V, et al. Functional Impairment of Microglia Coincides with Beta-Amyloid Deposition in Mice with Alzheimer-Like Pathology. PLoS One. 2013;8(4):e60921. doi:10.1371/journal.pone.0060921
30. Dunn N, Mullee M, Perry VH, Holmes C. Association between Dementia and Infectious Disease. Alzheimer Dis Assoc Disord. 2005;19(2):91-94. doi:10.1097/01.wad.0000165511.52746
.1f
31. Jonsson T, Atwal JK, Steinberg S, et al. A mutation in APP protects against Alzheimer’s disease and age-related cognitive decline. Nature. 2012;488(7409):96-99. doi:10.1038/nature11283
32. Novikova G, Kapoor M, TCW J, et al. Integration of Alzheimer’s disease genetics and myeloid genomics identifies disease risk regulatory elements and genes. Nat Commun. 2021;12(1):1610. doi:10.1038/s41467-021-21823-y
33. Wightman DP, Jansen IE, Savage JE, et al. A genome-wide association study with 1,126,563 individuals identifies new risk loci for Alzheimer’s disease. Nat Genet. 2021;53(9):1276-1282. doi:10.1038/s41588-021-00921-z
34. Holstege H, Hulsman M, Charbonnier C, et al. Exome sequencing identifies rare damaging variants in ATP8B4 and ABCA1 as risk factors for Alzheimer’s disease. Nat Genet. 2022;54(12):1786-1794. doi:10.1038/s41588-022-01208-7
35. Ozben T, Ozben S. Neuro-inflammation and anti-inflammatory treatment options for Alzheimer’s disease. Clin Biochem. 2019;72:87-89. doi:10.1016/j.clinbiochem.2019.04.001
36. Akiyama H, Barger S, Barnum S, et al. Inflammation and Alzheimer’s disease. Neurobiol Aging. 2000;21(3):383-421. doi:10.1016/s0197-4580(00)00124-x
37. Ahmad MH, Fatima M, Mondal AC. Influence of microglia and astrocyte activation in the neuroinflammatory pathogenesis of Alzheimer’s disease: Rational insights for the therapeutic approaches. J Clin Neurosci. 2019;59:6-11. doi:10.1016/j.jocn.2018.10.034
38. Liu L, Chan C. The role of inflammasome in Alzheimer’s disease. Ageing Res Rev. 2014;15:6-15. doi:10.1016/j.arr.2013.12.007
39. McLarnon JG. A Leaky Blood-Brain Barrier to Fibrinogen Contributes to Oxidative Damage in Alzheimer’s Disease. Antioxidants (Basel). 2021;11(1). doi:10.3390/antiox11010102
40. Hanslik KL, Ulland TK. The Role of Microglia and the Nlrp3 Inflammasome in Alzheimer’s Disease. Front Neurol. 2020;11:570711. doi:10.3389/fneur.2020.570711
41. Ryu JK, McLarnon JG. A leaky blood-brain barrier, fibrinogen infiltration and microglial reactivity in inflamed Alzheimer’s disease brain. J Cell Mol Med. 2009;13(9a):2911- 2925. doi:10.1111/j.1582- 4934.2008.00434.x
42. Nation DA, Sweeney MD, Montagne A, et al. Blood–brain barrier breakdown is an early biomarker of human cognitive dysfunction. Nat Med. 2019;25(2):270-276. doi:10.1038/s41591-018-0297-y
43.Montagne A, Barnes SR, Sweeney MD, et al. Blood-Brain Barrier Breakdown in the Aging Human Hippocampus. Neuron. 2015;85(2):296-302. doi:10.1016/j.neuron.2014.12.032
44. Martinon F, Mayor A, Tschopp J. The inflammasomes: guardians of the body. Annu Rev Immunol. 2009;27:229-265. doi:10.1146/annurev.immunol.021908.13 2715
45. Minkiewicz J, de Rivero Vaccari JP, Keane RW. Human astrocytes express a novel NLRP2 inflammasome. Glia. 2013;61(7):1113-1121. doi:10.1002/glia.22499
46. Dinarello CA. IL-1: Discoveries, controversies and future directions. Eur J Immunol. 2010;40(3):599-606. doi:10.1002/eji.201040319
47. Ojala J, Alafuzoff I, Herukka SK, van Groen T, Tanila H, Pirttilä T. Expression of interleukin-18 is increased in the brains of Alzheimer’s disease patients. Neurobiol Aging. 2009;30(2):198-209. doi:10.1016/j.neurobiolaging.2007.06.00 6
48. Malaguarnera L, Motta M, Di Rosa M, Anzaldi M, Malaguarnera M. Interleukin-18 and transforming growth factor-beta 1 plasma levels in Alzheimer’s disease and vascular dementia. Neuropathology. 2006;26(4):307-312. doi:10.1111/j.1440-1789.2006.00701.x
49. OZTÜrk C, Ozge A, Yalin OO, et al. The diagnostic role of serum inflammatory and soluble proteins on dementia subtypes: correlation with cognitive and functional decline. Behavioural neurology. 2007;18(4):207-215. doi:10.1155/2007/432190
50. Rubio-Perez JM, Morillas-Ruiz JM. A review: inflammatory process in Alzheimer’s disease, role of cytokines. ScientificWorldJournal. 2012;2012:756357. doi:10.1100/2012/756357
51. Bagyinszky E, Giau V Van, Shim K, Suk K, An SSA, Kim S. Role of inflammatory molecules in the Alzheimer’s disease progression and diagnosis. J Neurol Sci. 2017;376:242-254. doi:10.1016/j.jns.2017.03.031
52. Kitazawa M, Cheng D, Tsukamoto MR, et al. Blocking IL-1 Signaling Rescues Cognition, Attenuates Tau Pathology, and Restores Neuronal β-Catenin Pathway Function in an Alzheimer’s Disease Model. The Journal of Immunology. 2011;187(12):6539-6549. doi:10.4049/jimmunol.1100620
53. Spooren A, Kolmus K, Laureys G, et al. Interleukin-6, a mental cytokine. Brain Res Rev. 2011;67(1-2):157-183. doi:10.1016/j.brainresrev.2011.01.002
54. Ryu JK, Rafalski VA, Meyer-Franke A, et al. Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration. Nat Immunol. 2018;19(11):1212-1223. doi:10.1038/s41590-018-0232-x
55.Di Filippo M, Chiasserini D, Tozzi A, Picconi B, Calabresi P. Mitochondria and the link between neuroinflammation and neurodegeneration. J Alzheimers Dis. 2010;20 Suppl 2:S369-79. doi:10.3233/JAD-2010-100543
56. Reddy PH. Amyloid beta, mitochondrial structural and functional dynamics in Alzheimer’s disease. Exp Neurol. 2009;218(2):286-292. doi:10.1016/j.expneurol.2009.03.042
57. Roser M. Artificial intelligence is transforming our world — it is on all of us to make sure that it goes well. Published December 15, 2020. Accessed April 6, 2023. https://ourworldindata.org/ai- impact
58. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Med. 2019;17(1):195. doi:10.1186/s12916-019-1426-2
59. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):94-98. doi:10.7861/futurehosp.6-2-94
60. Chaves R, Górriz JM, Ramírez J, Illán IA, Salas-Gonzalez D, Gómez-Río M. Efficient mining of association rules for the early diagnosis of Alzheimer’s disease. Phys Med Biol. 2011;56(18):6047-6063. doi:10.1088/0031-9155/56/18/017
61. Happawana KA, Diamond BJ. Association rule learning in neuropsychological data analysis for Alzheimer’s disease. J Neuropsychol. 2022;16(1):116-130. doi:10.1111/jnp.12252
62. Szalkai B, Grolmusz VK, Grolmusz VI, Coalition Against Major Diseases. Identifying combinatorial biomarkers by association rule mining in the CAMD Alzheimer’s database. Arch Gerontol Geriatr. 2017;73:300-307. doi:10.1016/j.archger.2017.08.006
63. Castillo-Barnes D, Su L, Ramírez J, et al. Autosomal Dominantly Inherited Alzheimer Disease: Analysis of genetic subgroups by Machine Learning. Inf Fusion. 2020;58:153- 167. doi:10.1016/j.inffus.2020.01.001
64. Kavitha C, Mani V, Srividhya SR, Khalaf OI, Tavera Romero CA. Early-Stage Alzheimer’s Disease Prediction Using Machine Learning Models. Front Public Health. 2022;10:853294. doi:10.3389/fpubh.2022.853294
65. Gaetani L, Bellomo G, Parnetti L, Blennow K, Zetterberg H, Di Filippo M. Neuroinflammation and Alzheimer’s Disease: A Machine Learning Approach to CSF Proteomics. Cells. 2021;10(8). doi:10.3390/cells10081930
66. Cheung CY, Ran AR, Wang S, et al. A deep learning model for detection of Alzheimer’s disease based on retinal photographs: a retrospective, multicentre case-control study. Lancet Digit Health. 2022;4(11):e806-e815. doi:10.1016/S2589-7500(22)00169-8
67. Venugopalan J, Tong L, Hassanzadeh HR, Wang MD. Multimodal deep learning models for early detection of Alzheimer’s disease stage. Sci Rep. 2021;11(1):3254. doi:10.1038/s41598-020-74399-w
68. Pan D, Zeng A, Jia L, Huang Y, Frizzell T, Song X. Early Detection of Alzheimer’s Disease Using Magnetic Resonance Imaging: A Novel Approach Combining Convolutional Neural Networks and Ensemble Learning. Front Neurosci. 2020;14. doi:10.3389/fnins.2020.00259
69. Shastry KA, Vijayakumar V, Manoj Kumar M V, Manjunatha BA, Chandrashekhar BN. Deep Learning Techniques for the Effective Prediction of Alzheimer’s Disease: A Comprehensive Review. Healthcare (Basel). 2022;10(10):1842. doi:10.3390/healthcare10101842
70. Bi X, Li S, Xiao B, Li Y, Wang G, Ma X. Computer aided Alzheimers disease diagnosis by an unsupervised deep learning technology. Neurocomputing. 2020;392:296-304. doi:https://doi.org/10.1016/j.neucom.201 8.11.111
71. Liu Y, Mazumdar S, Bath PA. An unsupervised learning approach to diagnosing Alzheimer’s disease using brain magnetic resonance imaging scans. Int J Med Inform. 2023;173:105027. doi:https://doi.org/10.1016/j.ijmedinf.20 23.105027
72. Alashwal H, El Halaby M, Crouse JJ, Abdalla A, Moustafa AA. The Application of Unsupervised Clustering Methods to Alzheimer’s Disease. Front Comput Neurosci. 2019;13:31. doi:10.3389/fncom.2019.00031
73. Wang M, Roussos P, McKenzie A, et al. Integrative network analysis of nineteen brain regions identifies molecular signatures and networks underlying selective regional vulnerability to Alzheimer’s disease. Genome Med.2016;8(1):104. doi:10.1186/s13073-016-0355-3
74. Canario E, Chen D, Han Y, Niu H, Biswal B. Global Network Analysis of Alzheimer’s Disease with Minimum Spanning Trees. J Alzheimers Dis. 2022;89(2):571-581. doi:10.3233/JAD-215573
Author Area
Have an article to submit?
Submission Guidelines
Submit a manuscript
Become a member