Giorgi Svanishvili 
Anatomical Researches and Skills Centre, Tbilisi, Georgia
Correspondence to: giorgisvanishvili85@gmail.com
Additional information
- Ethical approval: N/a
- Consent: N/a
- Funding: No industry funding
- Conflicts of interest: N/a
- Author contribution: Giorgi Svanishvili – Conceptualization, Writing – original draft, review and editing
- Guarantor: Giorgi Svanishvili
- Provenance and peer-review:
Commissioned and externally peer-reviewed - Data availability statement: N/a
Keywords: AI in Alzheimer’s diagnostics, Early Alzheimer’s detection, Blood biomarkers in AD, Machine learning in neuroimaging, Personalized Alzheimer’s care.
Peer Review
Received: 22 December 2024
Revised: 2 January 2025
Accepted: 2 January 2025
Published: 18 January 2025
Abstract
Artificial intelligence (AI) is changing how Alzheimer’s disease (AD) is diagnosed and managed. AI tools can analyze simple tests, like blood samples, to detect early signs of Alzheimer’s with high accuracy, often before symptoms appear. For example, studies have shown that these tools can identify Alzheimer’s with over 85% accuracy by analyzing specific changes in blood markers, such as reduced levels of d-glutamate, which are linked to memory and cognitive decline. AI can also process brain scans to track changes, such as shrinking memory-related areas of the brain, achieving over 90% accuracy in identifying early stages of the disease. These advancements make diagnosis faster, less invasive, and more affordable than traditional methods like spinal fluid tests or specialized brain scans, which are often costly and difficult to access. However, challenges remain, including the need for more diverse data to improve reliability and make these tools widely available in clinics. With further development, AI has the potential to improve early diagnosis, provide personalized care, and significantly reduce the global impact of AD on patients and families.
Introduction
Alzheimer’s disease (AD) is the leading cause of dementia worldwide, with a global prevalence estimated at approximately 50 million people in 2018. This number is projected to triple by 2050, largely driven by aging populations, with two-thirds of cases occurring in low- and middle-income countries.1 In Europe alone, dementia prevalence is expected to double by 2050.2 While some evidence suggests that the incidence of dementia may be declining in high-income countries,3 this trend does not hold uniformly, and the overall prevalence remains a significant challenge.4 AD imposes a heavy burden not only on individuals and their caregivers but also on healthcare systems due to its high costs and resource demands. A US-based study involving nearly 60,000 individuals reported survival times of 3–4 years.5 In contrast, a European memory-clinic-based cohort showed a median survival time of 6 years.6 Furthermore, estimates suggest that the total disease duration can span up to 2 decades, with approximately 10 years in the preclinical stage, 4 years in the prodromal stage (characterized by mild cognitive impairment (MCI)), and 6 years in the dementia stage. Importantly, a biologically defined prevalence of AD at age 85 is three times higher than that defined clinically, underscoring the need for better diagnostic methods that account for early, subclinical stages.7
Significant progress has been made in understanding Alzheimer’s pathology, identifying causative and protective genetic factors, and developing biomarkers for early diagnosis. However, recent controversies have raised concerns about the integrity of some foundational research. In particular, allegations of manipulated data in a 2006 Nature paper have cast doubt on conclusions about a specific amyloid-beta (Abeta) aggregate (Abeta*56) as the key toxic factor in AD.8 While this incident underscores the challenges in scientific research, the amyloid hypothesis remains central to the field, supported by thousands of independent studies. Nevertheless, alternative hypotheses and combination treatment strategies continue to be explored, emphasizing the need for robust, reproducible evidence in Alzheimer’s research. This review examines the current state of Alzheimer’s diagnostics, highlighting the limitations of traditional methods and the promising role of artificial intelligence (AI) in improving accuracy, enabling earlier intervention, and addressing unmet needs in clinical practice.
Alzheimer’s Disease
Alzheimer’s disease is a progressive neurodegenerative disorder and the most common cause of dementia, characterized by substantial cognitive and functional decline. Its clinical presentations vary, ranging from memory-dominant amnestic variants, typically seen in patients over 70 years old, to less common language-centric or atypical variants in younger individuals under 70. These variations complicate timely recognition and diagnosis, particularly when memory loss is not the most prominent symptom.9 The pathological hallmarks of AD include the accumulation of amyloid-beta plaques and tau protein tangles in the brain. Amyloid-beta is a product of the breakdown of the amyloid precursor protein (APP), a process that occurs throughout life. While APP turnover happens within hours, clearance of amyloid-beta slows with aging. This leads to the formation of misfolded, less-soluble fragments that aggregate into plaques, a defining feature of Alzheimer’s pathology. These plaques, alongside tau tangles, trigger neuroinflammation and contribute to the disease’s progression.10
Despite its importance, the amyloid hypothesis is not without controversy. For instance, while anti- amyloid therapies have demonstrated success in removing amyloid plaques, their ability to halt or reverse cognitive decline remains unproven. Moreover, amyloid accumulation alone is insufficient to cause AD; it acts as a necessary but not exclusive factor. Other contributing mechanisms, such as inflammation and vascular pathology, underline the multifactorial nature of the disease.11 Advancements in diagnostic approaches have shifted the focus from clinical symptomatology to biomarker- based frameworks. The ATN (amyloid, tau, neurodegeneration) framework categorizes AD based on the presence of these biomarkers, enabling diagnoses at preclinical or MCI stages (Figure 1). However, the framework is not without its limitations. Operational challenges, including undefined cutoff points for biomarkers and exclusion of other dementia-related pathologies, restrict its clinical applicability.
Source: Rik Ossenkoppele and Gil Rabinovici https://doi.org/10.1016/ S0140-6736(20)32205-4.
ADappt (Alzheimer’s Diagnostic Assessment and Prediction Tool) and other tools are being developed to bridge gaps in diagnosis and improve clinician-patient communication, particularly in the early stages.12 Yet, challenges remain in providing individualized prognoses, especially for patients with subjective cognitive decline. The pressing need for improved diagnostic accuracy and early detection highlights the importance of novel approaches, such as AI, to address these gaps and revolutionize the diagnostic process for AD.
Traditional Diagnostics of AD
Historically, AD was diagnosed as an exclusionary condition, with most diagnoses made in the late stages of the disease when cognitive decline and functional impairment were obvious.13 This late diagnosis has serious consequences, as AD progresses over decades, taking a heavy toll on patients, caregivers, and healthcare systems. In most cases, patients with early signs or symptoms of AD initially present in a primary care setting. Subtle cognitive or behavioral changes might be observed during routine wellness visits or when addressing unrelated comorbidities. However, detecting these early effects is not straightforward, as neuropathological hallmarks of AD—amyloid-beta (Aβ) plaques and neurofibrillary tangles—can develop decades before clinical symptoms emerge.14,15 (Figure 2).
Source: https://doi.org/10.14283/jpad.2021.23.
Advances in biomarker research have transformed the AD diagnostic landscape, providing techniques for detecting pathogenic alterations early in the disease progression. Imaging methods such as magnetic resonance imaging (MRI) and positron emission tomography (PET) enable visualization of structural and molecular changes in the brain.16 Amyloid PET scans, in particular, use tracers (e.g., florbetapir, flutemetamol, and florbetaben) to bind to Aβ within plaques, allowing clinicians to quantify amyloid pathology directly.17 Alternatively, cerebrospinal fluid (CSF) biomarkers offer another robust method for detecting AD-related pathology. Lumbar punctures enable the collection of CSF, which can be analyzed for key biomarkers such as decreased Aβ42 (indicative of Aβ aggregation) and increased phosphorylated tau (p-tau) and total tau (t-tau), reflecting tau pathology and neurodegeneration, respectively.18 Ratios of biomarkers, such as Aβ42/Aβ40, provide a more precise measure of amyloid-related pathology.19 These indicators can be recognized long before obvious clinical signs appear, making them useful for early-stage diagnosis.
Despite significant advancements, traditional diagnostic approaches face several challenges. Amyloid PET and CSF biomarker testing are not universally accessible due to financial, practical, and regional disparities in healthcare systems. Furthermore, neither method alone can definitively diagnose clinical AD, as results must be interpreted alongside cognitive assessments and other clinical findings. Additionally, both amyloid and tau pathologies are present in other neurological conditions, complicating differential diagnosis. Traditional methods have laid the foundation for understanding and diagnosing AD, but their limitations underscore the urgent need for innovative approaches. These methods, while effective in supporting diagnoses, are hindered by practical and systemic challenges. Emerging technologies, such as AI, may offer solutions to overcome these barriers and enhance the precision, accessibility, and timeliness of AD diagnosis.
AI and AD
Artificial intelligence, particularly machine learning (ML) and deep learning (DL), has emerged as a transformative tool in AD research (Figure 3). These technologies are enhancing early detection, monitoring disease progression, and enabling personalized care by leveraging complex data from imaging, biomarkers, and genetic information.
AI-driven algorithms are proving invaluable in identifying early signs of AD, often before clinical symptoms manifest. ML models, trained on MRI scans, can detect subtle structural changes in the brain indicative of AD. These models analyze intricate patterns that are imperceptible to the human eye, advancing the possibility of pre-symptomatic diagnosis. For example, DL models utilizing convolutional neural networks (CNNs) have demonstrated exceptional accuracy in classifying images, with layers extracting features like edges, textures, and patterns that may signal early disease progression.20 CNNs process images layer by layer, beginning with simple features such as edges and progressing to more complex patterns, similar to how the human brain processes visual
information.
Beyond MRI, PET imaging has benefited from AI advancements. Studies utilizing deep CNNs (Figure 4), such as Choi et al.’s work on FDG and florbetapir PET scans, have shown predictive accuracies of over 84% for MCI progressing to AD.21 These CNNs eliminate the need for manually defined feature extraction, using minimally processed 3D PET images to predict cognitive decline and longitudinal changes. DL algorithms were utilized to distinguish between individuals with AD and healthy controls using tau PET scans. This 3D CNN-based classification model achieved an average accuracy of 90.8% using five-fold cross- validation. In addition, the researchers employed a layer-wise relevance propagation model to determine which brain regions in tau PET pictures contributed the most to the categorization results. The most commonly identified brain regions were the hippocampus, parahippocampus, thalamus, and fusiform gyrus.22
Source: https://doi.org/10.1016/j.ejrad.2023.110934.
AI’s versatility extends to predictive modeling, where algorithms integrate diverse data sources such as genetic, cognitive, and lifestyle factors to assess individual AD risk and progression. These models provide clinicians with prognostic insights, enabling tailored treatment and care strategies. For instance, Qiu et al.’s DL framework for AD classification, validated across multiple datasets, achieved near-perfect accuracy (area under curve values up to 0.996) and outperformed practicing neurologists. Such models link imaging signatures to AD pathophysiology, offering a clinically adaptable approach for routine diagnosis.23
Studies have highlighted the importance of glutamate and d-amino acids as potential biomarkers. N-methyl-d-aspartate (NMDA) receptors, crucial for cognitive functions, mediate glutamate signaling, which plays a fundamental role in learning and memory.24 Disruptions in this pathway have been implicated in AD. Recent research shows that plasma d-glutamate levels are significantly reduced in AD patients compared to healthy controls, correlating with cognitive decline as measured by tools like the Mini-Mental State Examination (MMSE).25 A study involving 397 participants found that plasma d-glutamate levels in MCI and AD patients were markedly lower. Furthermore, the MMSE scores were significantly correlated with d-glutamate levels (adjusted R² = 0.344), reinforcing its potential as a biomarker for cognitive decline.
ML models have effectively integrated these biomarkers for diagnostic purposes.26 For instance, Chang et al. employed ML algorithms, including random forests (RFs) and naïve Bayes models, to distinguish MCI and AD patients from healthy individuals using plasma d-glutamate levels. Their naïve Bayes model achieved an AUC of 0.8207, with sensitivity and specificity rates of 84.4% and 81.6%, respectively.27 AUC, or Area Under the Curve, is a measure used to evaluate the performance of diagnostic models. It reflects how well a model distinguishes between two groups, such as individuals with and without Alzheimer’s. AUC values range from 0.5 (random guessing) to 1.0 (perfect classification), with higher values indicating better accuracy. For instance, an AUC of 0.85 means the model can correctly identify cases 85% of the time.
However, evaluating d-glutamate for clinical usage presents hurdles, such as the role of altered glutamate availability and NMDA receptor regulation in AD pathogenesis. According to research,28 poor glutamate recycling and transporter function in AD may raise extracellular glutamate levels, worsening excitotoxicity, and neurodegeneration. Amyloid-beta peptides can directly modify NMDA receptor function, leading to increased Ca+ influx and synaptic toxicity.28 Understanding these mechanisms is critical for improving d-glutamate’s significance as a biomarker and resolving fluctuation in its levels due to complicated illness interactions. Correlation Explanation (CorEx), a novel unsupervised ML technique, has demonstrated the ability to predict AD by analyzing complex interactions among imaging, genetic, and biomarker data. In a study involving 829 participants, CorEx identified latent factors associated with cognitive decline and brain atrophy, achieving a prediction accuracy of nearly 90% on independent datasets. These factors, including markers of cardiovascular, immune, and bioenergetic functions, align with known AD pathways, underscoring AI’s capacity to uncover novel biological insights.29
Recently, pilot studies have investigated blood metabolites as potential indicators for AD.30 Stamate and his colleagues30,31 analyzed data from the European Medical Information Framework for AD Multimodal Biomarker Discovery.32 Blood-based metabolites, a less invasive alternative to CSF biomarkers, have been explored using ML algorithms like extreme gradient boosting (XGBoost) and RFs. These models differentiate AD patients from cognitively normal individuals with AUC values exceeding 0.85, demonstrating the feasibility of blood-based diagnostics. Additionally, AI has been instrumental in interpreting diffusion tensor imaging (DTI) data to assess white matter integrity in predementia stages. Studies utilizing support vector machines with DTI metrics, such as fractional anisotropy and mean diffusivity, have shown higher prediction accuracies than traditional gray matter volume analysis, particularly in distinguishing MCI patients with amyloid-beta positivity.33 Results considered ML with novel biomarkers and multiple variables may increase the sensitivity and specificity in the diagnosis of AD.34 Rapid and cost-effective high-performance liquid chromatography (HPLC) for biomarkers and ML algorithms may assist physicians in diagnosing AD in outpatient clinics.
Potential Therapies
AI-powered tools could make routine diagnostics faster and more accessible. For example, outpatient clinics might use ML models with HPLC to analyze blood samples for biomarkers like d-glutamate. Instead of waiting for expensive PET imaging or invasive CSF analysis, a clinician could confirm early signs of Alzheimer’s within minutes. This approach is rapid, affordable, and scalable, ensuring accurate diagnoses even in regions without advanced facilities. Imagine a patient visiting their local clinic for forgetfulness. A simple blood draw analyzed by AI could flag a potential risk for Alzheimer’s long before cognitive symptoms escalate.
AI models analyzing sequential MRI or PET scans could allow doctors to detect subtle changes in brain structure, like hippocampal shrinkage or tau buildup. These changes, tracked over time, would provide a detailed picture of how quickly the disease is progressing for each patient. For instance, a clinician could use AI to determine that a patient’s tau accumulation is advancing at a slower rate than expected, prompting adjustments to their treatment regimen. Patients could also benefit from wearable devices linked to AI systems. These devices might monitor cognitive or physical performance, sending real-time updates to physicians and caregivers. This could reduce unnecessary doctor visits while ensuring timely interventions when significant changes occur. AI could help healthcare systems save billions by diagnosing AD earlier, reducing the reliance on long-term care. For example, an at-risk individual identified through AI-driven blood biomarker analysis might begin preventive treatments years before requiring intensive care.
Limitations
While AI and related technologies show immense promise in revolutionizing AD diagnostics and management, several limitations exist within the studies and approaches discussed, both technically and practically. Most AI models, including those trained on imaging data like MRI and PET scans, rely on datasets that may not represent the diversity of the global population. For instance, studies often use datasets such as the Alzheimer’s Disease Neuroimaging Initiative, which predominantly includes individuals from high-income countries. These datasets may not account for variations in genetics, lifestyle, or access to healthcare seen in other populations, reducing the models’ applicability in low- and middle-income regions.
The effectiveness of AI models depends on the quality and completeness of input data. Imaging studies, for instance, require high-resolution, standardized scans, which may not always be available in real-world clinical settings. Similarly, blood and CSF biomarker studies rely on accurate and consistent sample collection methods, which can vary across laboratories and regions. While CSF analysis is highly effective in detecting biomarkers like amyloid-beta and tau, it remains invasive and uncomfortable for patients. Even though blood-based biomarker studies are promising, they are still in the early stages and require further validation before being widely adopted. Furthermore, some biomarkers, like d-glutamate, have shown correlations with cognitive decline, but the underlying mechanisms are not yet fully understood, limiting their current clinical utility.
Implementing AI in routine clinical workflows involves significant computational demands. Hospitals and clinics in resource-limited settings may lack the infrastructure to process large datasets or run complex algorithms. Moreover, integrating AI systems with existing electronic health records requires standardization and coordination, which can be time-consuming and expensive. In addition to technical and practical challenges, ethical concerns about data privacy and security are important barriers to wider adoption. The delicate nature of patient data, particularly medical imaging and genetic information, necessitates strict security measures to prevent unauthorized access or use. Any lapses in data security could result in a loss of patient trust and even in legal consequences.
Conclusion
Artificial intelligence transforms AD care by enabling earlier diagnosis, personalized treatment, and more accessible diagnostic methods. Tools like blood-based biomarker analysis and imaging models have demonstrated high accuracy and scalability, addressing the limitations of traditional approaches. However, challenges remain, including limited generalizability of AI models, the need for better validation of emerging biomarkers like d-glutamate, and infrastructure barriers in clinical settings. Addressing these issues is critical to realizing AI’s full potential. To address these challenges, future efforts should focus on data diversity by including datasets from underrepresented populations, as well as ensuring AI models are adaptable across demographics and healthcare systems. Furthermore, the development of low-cost diagnostic tools customized for low-income and resource-constrained locations is critical to closing healthcare gaps and making AI-driven solutions universally accessible. With further research and integration, AI can significantly improve outcomes for patients and reduce the global burden of AD.
Reference
1 Alzheimer’s Disease International. World Alzheimer Report 2018. The state of the art of dementia research: new frontiers. September, 2018. Available from: https://www.alzint.org/u/WorldAlzheimerReport2018.pdf (accessed Sept 9, 2020).
2 Alzheimer Europe. Dementia in Europe Yearbook 2019: Estimating the prevalence of dementia in Europe. 2020. Available from: https://www.alzheimereurope.org/content/download/195515/1457520/file/FINAL %20 05707%20Alzheimer%20Europe%20yearbook%202019.pdf (accessed Jan 24, 2021).
3 Wu YT, Beiser AS, Breteler MMB, Fratiglioni L, Helmer C, Hendrie HC, et al. The changing prevalence and incidence of dementia over time – current evidence. Nat Rev Neurol. 2017;13(6):327-39.
https://doi.org/10.1038/nrneurol.2017.63
4 Prince M, Ali GC, Guerchet M, Prina AM, Albanese E, Wu YT. Recent global trends in the prevalence and incidence of dementia, and survival with dementia. Alzheimers Res Ther. 2016;8(1):23.
https://doi.org/10.1186/s13195-016-0188-8
5 Mayeda ER, Glymour MM, Quesenberry CP, Johnson JK, Pérez-Stable EJ, Whitmer RA. Survival after dementia diagnosis in five racial/ethnic groups. Alzheimers Dement. 2017;13(7):761-9.
https://doi.org/10.1016/j.jalz.2016.12.008
6 Rhodius-Meester HFM, Tijms BM, Lemstra AW, Prins ND, Pijnenburg YAL, Bouwman F, et al. Survival in memory clinic cohort is short, even in young-onset dementia. J Neurol Neurosurg Psychiatry. 2019;90(6):726-8.
https://doi.org/10.1136/jnnp-2018-318820
7 Jack CR, Therneau TM, Weigand SD, Wiste HJ, Knopman DS, Vemuri P, et al. Prevalence of biologically vs clinically defined Alzheimer spectrum entities using the National Institute on Aging-Alzheimer’s association research framework. JAMA Neurol. 2019;76(10):1174-83.
https://doi.org/10.1001/jamaneurol.2019.1971
8 Turner RS. Allegations of fraud in Alzheimer’s disease research: Death of the amyloid hypothesis? Georgetown University; [cited 2024 Dec 11]. Available from: https://memory.georgetown.edu/news/allegations-of-fraud-in-alzheimers-disease-research-death-of-the-amyloid-hypothesis%EF%BF%BC/
9 Scheltens P, De Strooper B, Kivipelto M, Holstege H, Chételat G, Teunissen CE, et al. Alzheimer’s disease. Lancet. 397(10284):1577-90
https://doi.org/10.1016/S0140-6736(20)32205-4
10 Jack CR Jr, Holtzman DM, Sperling R. Dementia is not synonymous with Alzheimer’s disease. Sci Transl Med. 2019;11:eaav0511.
https://doi.org/10.1126/scitranslmed.aav0511
11 Sweeney MD, Montagne A, Sagare AP, Nation DA, Schneider LS, Chui HC, et al. Vascular dysfunction-The disregarded partner of Alzheimer’s disease. Alzheimers Dement. 2019;15(1):158-67.
https://doi.org/10.1016/j.jalz.2018.07.222
12 van Maurik IS, Visser LN, Pel-Littel RE, van Buchem MM, Zwan MD, Kunneman M, et al. Development and usability of ADappt:
Web-based tool to support clinicians, patients, and caregivers in the diagnosis of mild cognitive impairment and Alzheimer disease. JMIR Form Res. 2019;3:e13417.
https://doi.org/10.2196/13417
13 Sabbagh MN, Lue LF, Fayard D, Shi J. Increasing precision of clinical diagnosis of Alzheimer’s disease using a combined algorithm incorporating clinical and novel biomarker data. Neurol Ther. 2017;6(Suppl 1):83-95.
https://doi.org/10.1007/s40120-017-0069-5
14 Bateman RJ, Xiong C, Benzinger TL, Fagan AM, Goate A, Fox NC, et al. Clinical and biomarker changes in dominantly inherited Alzheimer’s disease. N Engl J Med. 2012;367(9):795-804.
https://doi.org/10.1056/NEJMoa1202753
15 Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018;14(4):535-62.
https://doi.org/10.1016/j.jalz.2018.02.018
16 Liu JL, Hlavka JP, Hillestad R, Mattke S. Assessing the preparedness of the U.S. Health Care System infrastructure for an Alzheimer’s treatment 2017. Available from: https://www.rand.org/pubs/research_reports/RR2272.html (accessed May 5, 2018).
https://doi.org/10.7249/RR2272
17 Villemagne VL, Doré V, Burnham SC, Masters CL, Rowe CC. Imaging tau and amyloid-β proteinopathies in Alzheimer disease and other conditions. Nat Rev Neurol. 2018;14(4):225-36.
https://doi.org/10.1038/nrneurol.2018.9
18 Blennow K, Dubois B, Fagan AM, Lewczuk P, de Leon MJ, Hampel H. Clinical utility of cerebrospinal fluid biomarkers in the diagnosis of early Alzheimer’s disease. Alzheimers Dement. 2015;11(1):58-69.
https://doi.org/10.1016/j.jalz.2014.02.004
19 Hansson O, Lehmann S, Otto M, Zetterberg H, Lewczuk P. Advantages and disadvantages of the use of the CSF amyloid β (Aβ) 42/40 ratio in the diagnosis of Alzheimer’s disease. Alzheimers
Res Ther. 2019;11(1):34.
https://doi.org/10.1557/jmr.2019.190
20 Teuwen J, Moriakov N. Handbook of Medical Image Computing and Computer Assisted Intervention. 2020;(pp. 481-501). Elsevier
https://doi.org/10.1016/B978-0-12-816176-0.00025-9
21 Choi H, Jin KH. Predicting cognitive decline with deep learning of brain metabolism and amyloid imaging. Behav Brain Res. 2018;344:103-9.
https://doi.org/10.1016/j.bbr.2018.02.017
22 Jo T, Nho K, Risacher SL, Saykin AJ. Deep learning detection of
informative features in tau PET for Alzheimer’s disease classification. BMC Bioinformatics. 2020;21:496.
https://doi.org/10.1186/s12859-020-03848-0
23 Qiu S, Joshi PS, Miller MI, Xue C, Zhou X, Karjadi C, et al. Development and validation of an interpretable deep learning framework for Alzheimer’s disease classification. Brain. 1920;143:1920-33.
https://doi.org/10.1093/brain/awaa137
24 Li F, Tsien JZ. Memory and the NMDA receptors. N Engl J Med. 2009; 361:302-3.
https://doi.org/10.1056/NEJMcibr0902052
25 Lin CH, Yang HT, Lane HY. D-glutamate, D-serine, and D-alanine differ in their roles in cognitive decline in patients with Alzheimer’s disease or mild cognitive impairment. Pharmacol Biochem Behav. 2019;185:172760.
https://doi.org/10.1016/j.pbb.2019.172760
26 Lin CH, Yang HT, Chiu CC, Lane HY. Blood levels of d-amino acid oxidase vs. d-amino acids in reflecting cognitive aging. Sci. Rep. 2017;7:14849.
https://doi.org/10.1038/s41598-017-13951-7
27 Chang CH, Lin CH, Liu CY, Huang CS, Chen SJ, Lin WC, et al. Plasma D-glutamate levels for detecting mild cognitive impairment and
Alzheimer’s disease: machine learning approaches. J Psychopharmacol. 2021;35:265-72.
https://doi.org/10.1177/0269881120972331
28 Wang R, Reddy PH. Role of glutamate and NMDA receptors in Alzheimer’s disease. J Alzheimers Dis. 2017;57(4):1041-8. doi: 10.3233/JAD-160763.
https://doi.org/10.3233/JAD-160763
29 Riedel BC, Daianu M, Ver Steeg G, Mezher A, Salminen LE,
Galstyan A, et al. Uncovering biologically coherent peripheral signatures of health and risk for Alzheimer’s disease in the aging brain. Front Aging Neurosci. 2018;10:390.
https://doi.org/10.3389/fnagi.2018.00390
30 Varma VR, Oommen AM, Varma S, Casanova R, An Y, Andrews RM, et al. Brain and blood metabolite signatures of pathology and progression in Alzheimer disease: A targeted metabolomics study. PLoS Med. 2018;15:e1002482.
https://doi.org/10.1371/journal.pmed.1002482
31 Stamate D, Kim M, Proitsi P, Westwood S, Baird A, Nevado-Holgado A, et al. A metabolite-based machine learning approach to diagnose Alzheimer-type dementia in blood: Results from the European Medical Information Framework for Alzheimer Disease biomarker discovery cohort. Alzheimers Dement (N Y). 2019;5:933-8.
https://doi.org/10.1016/j.trci.2019.11.001
32 Kim M, Snowden S, Suvitaival T, Ali A, Merkler DJ, Ahmad T, et al. Primary fatty amides in plasma associated with brain amyloid burden, hippocampal volume, and memory in the European Medical Information Framework for Alzheimer’s Disease biomarker discovery cohort. Alzheimers Dement. 2019;15:817-27.
https://doi.org/10.1016/j.jalz.2019.03.004
33 Dyrba M, Barkhof F, Fellgiebel A, Filippi M, Hausner L, Hauenstein K, et al. Predicting prodromal Alzheimer’s disease in subjects with mild cognitive impairment using machine learning classification of multimodal multicenter diffusion-tensor and magnetic resonance imaging data. J Neuroimaging. 2015;25:738-47.
https://doi.org/10.1111/jon.12214
34 Chang C-H, Lin CH, Lane HY. Machine learning and novel biomarkers for the diagnosis of Alzheimer’s disease. Int J Mol Sci. 2021;22(5):2761. https://doi.org/10.3390/ijms22052761
https://doi.org/10.3390/ijms22052761
