Due to its comprehensive coverage of over 12,000 academic journals and its frequent use by researchers, Web of Science was selected as the primary database for this study. Compared with other databases such as Scopus, Medline, and PubMed, Web of Science provides the most comprehensive and reliable bibliometric analysis (14). The Web of Science Core Collections Database (WOSCC) was searched on September 8, 2024, for literature on artificial intelligence and Alzheimer’s disease published between January 1, 2010, and September 8, 2024. To reduce the bias caused by frequent database updates, we completed the search in 1 day. There was a search formula set as follows: [TS = (Alzheimer’s disease) or (Alzheimer) or (Alzheimer’s disease (AD)) or (AD) or (Alzheimer’s disease) or (Alzheimer Diseases) or (Alzheimer’s Diseases)] and [TS = (artificial intelligence) or (AI) or (artificial-intelligence)] and [LA = (English)]. In this study, only articles and reviews were considered (editorials, proceedings papers, abstracts, and book chapters were excluded), with a time frame from January 1, 2010, to September 8, 2024. Figure 1 illustrates how paper screening is conducted. Finally, in order to ensure data quality, we perform data cleansing, which includes the removal of duplicate records, uniform formatting of author names and dates, and handling of missing values. For records with critical information missing, we adopted a strategy of either filling in or excluding the records.

Bibliometric analysis often involves the application of mathematical and statistical methods to scrutinize research outcomes, with the goal of extracting meaningful information and identifying dominant patterns within extensive scholarly works (15, 16). This analytical approach offers insights into various bibliometric dimensions, including countries, institutions, journals, authors, keywords, and citations, presenting a holistic dataset encompassing these elements (17). Moreover, through visualization, it becomes possible to evaluate the ongoing research advancements in specific domains and to foresee emerging trends and focal points (18). In the field of bibliometrics, VOSviewer, CiteSpace, and Bibliometrix are three software tools designed for the analysis of scientific literature networks (19).

VOSviewer is a software tool for constructing and visualizing scientific literature networks (20). It processes citation, co-citation, co-word, and co-authorship data, generating intuitive scientific maps that help researchers identify research trends and knowledge structures (21). The software offers three types of visualizations: network view, overlay view, and density view (22). In our study, we conducted an in-depth bibliometric analysis using VOSviewer software. Initially, we set the analysis parameters based on our research objectives, including the time frame, node types, relationship strengths, threshold, clustering algorithm, clustering parameters, and visualization parameters to ensure the precision and relevance of the results. Subsequently, the software automatically constructed collaboration and co-citation networks based on these parameters, visually displaying the interrelationships and knowledge flow among literature. Finally, we generated various charts, including collaboration and co-citation network diagrams, which revealed the structure and dynamics within the research field. Through this process, we were able to clearly present and analyze key collaborations and knowledge development paths within the research area.

CiteSpace, based on the Java platform, it offers an intuitive approach to analyzing and visualizing large volumes of literature data, especially suitable for handling information from various academic database and helps researchers discover research hotspots and future trends in scientific literature (23, 24). In this study, we employed CiteSpace version 6.3.R1 to perform a comprehensive bibliometric analysis, which included the generation of clustering maps to visualize reference and keyword relationships and timeline visualizations to track the evolution of references over time. The software also enabled us to identify citation bursts within references, indicating significant research waves. Additionally, we conducted a dual-map analysis of journals to explore shifts in research directions and trends, revealing key themes and influential journals in the field. To achieve this, we set parameters such as time span, node types, relationship strengths, and co-occurrence thresholds, and leveraged CiteSpace’s clustering algorithms, including LSI, LLR, and MI, to identify research topics and trends. The software facilitated the extraction of terms from the titles, keywords, and abstracts of the literature, allowing us to construct collaboration and co-citation networks. These networks were then used to generate intuitive charts that illuminated the structure and dynamics of the research field, providing a clear overview of the key topics and influential publications.

Bibliometrix is an R package that covers the entire workflow of scientific mapping, from data collection to analysis and visualization (25). These tools enable researchers to identify research hotspots, trends, and patterns, providing valuable insights for academic studies. In addition, a bibliometric platform¹ is used to visualize international collaboration.

Following our search criteria, we identified a total of 530 research works related to artificial intelligence and Alzheimer’s disease since 2010, comprising 361 original articles and 169 review papers. There has been a consistent annual increase in the number of publications, with a particularly dramatic surge observed between 2019 and 2024. Even with only partial data for 2024, the count has already exceeded 133 articles, indicating a persistent upward trajectory. However, there was a negligible amount of related research prior to 2015 (Figure 2).

A total of 76 countries/regions participated in research on artificial intelligence and Alzheimer’s disease between 2010 and 2024. The top 10 countries/regions with significant contributions and notable associations are presented in Table 1, it can reflect the countries/regions that have made the most significant contributions in this area. Figure 3A and Table 1 show the top 10 countries and institutions according to publication number, and the top 5 countries are the United States, China, India, England and Italy. Table 1 indicates that the United States has a leading role with 152 publications, highlighting its significant contribution to the field. China follows with a notable 100 publications, indicating a strong research presence. India’s 60 publications reflect an emerging research force, while England’s 50 and Italy’s 44 demonstrate their active engagement in the area. In the analysis of citations per publication, Germany secured the top position with an average of 20.88 citations per publication, followed closely by Italy with 18.18 citations per publication. The United States and England made significant advancements, claiming the third and fourth spots with 17.86 and 17.04 citations per publication, respectively. Additionally, we examined the collaborative ties between nations, with lines connecting two countries signifying the presence of cooperative relationships, the analysis of the global country collaboration network illustrates the interconnectedness and collaborative efforts among various nations in the field. As illustrated in Figures 3B,C, the analysis revealed a pattern of intensified collaboration among a select group of countries. Specifically, the United States, the United Kingdom, Canada, India, China, and France exhibited a higher degree of cooperative engagement with one another. The strength of these connections implies a significant impact on the collective advancement of research and innovation in their respective fields of study.

It is estimated that over 1,228 institutions have contributed to the research on artificial intelligence and Alzheimer’s disease. As shown in Figure 4A, seven of the top 10 most productive institutions were from the United States, as Harvard University ranks first with 47 publications, followed by Indiana University System in second position with 42 articles. In addition, the institutions with the third highest number of publications are University of London from the United Kingdom, with 33 articles published. It is clear that both the United States and the United Kingdom have significant influence in this field. In addition, we performed a bibliographic coupling analysis to explore the intellectual and research connections among institutions with a minimum threshold of 5 publications, as depicted in Figure 4B, from the 1,228 institutions initially considered, 39 met the criteria for inclusion in the analysis. This method, which discerns relationships through the co-citation of documents, offers a graphical depiction of the inter-institutional scholarly links. Harvard Medical School emerged with the most substantial total link strength, totaling 4,319, succeeded by University of Cambridge with 4,235, and the Alan Turing Institute with 3,880. Additionally, the overlay map visualization indicates that research on artificial intelligence and Alzheimer’s disease has only rapidly developed in recent years, demonstrating the immense potential of artificial intelligence technology in the early diagnosis of Alzheimer’s disease, monitoring disease progression, developing treatment strategies, and researching pathological mechanisms in the future.

A total of 282 distinct journals have published 530 articles in the field of artificial intelligence and Alzheimer’s disease. Among these journals, five have over 10 papers published. Table 2 shows the top 10 journals/co-cited journals with the number of publications and citations/co-citations, and their corresponding Impact Factor (JCR 2023) and JCR quartile. As depicted in Figure 5A and Table 2, the top 10 journals with the highest number of published articles are showcased. This ranking reflects the prolific contribution of these journals to the field, indicating their significant role in disseminating research findings and advancing scientific knowledge. Notably, the Journal of Alzheimer’s Disease (IF = 3.4, Q2) stands out prominently in the field, contributing 27 publications and a substantial total link strength of 346, emphasizing its significant role in advancing research on Alzheimer’s disease. Closely trailing, the Diagnostics (IF = 3, Q1) exhibits a notable presence as well, publishing 15 articles and achieving a total link strength of 154. Furthermore, the Frontiers in Aging Neuroscience (IF = 4.1, Q2) has made a remarkable impact, published 15 articles, demonstrating an even higher total link strength of 434. These journals, frequently cited in the realm of Alzheimer’s disease research, underscore their influential positions not only in the broader neuroscience community but also within the intersection of artificial intelligence and Alzheimer’s disease studies.

Journals that have garnered citations from multiple authors are designated as co-cited journals. As depicted in Figure 5B, we conducted a visual analysis focusing on these co-cited journals that received over 50 citations. Among them, the Alzheimer’s & Dementia emerged as the top contender, boasting an Impact Factor of 13 and a remarkable 820 citations. This was closely followed by the Neuroimage, with an Impact Factor of 4.7 and 740 citations, and the Journal of Alzheimer’s Disease, holding an Impact Factor of 3.4 and 696 citations. These journals not only demonstrate their widespread influence within the field but also underscore their importance as key sources of information and research insights. Table 2 provides detailed information on journals and co-cited journals, which allows you to explore publication preferences among authors and provide useful references for future research.

Using the double-map overlay of journals shown in Figure 5C, we can see the left side displays the distribution of journals publishing the citing literature, reflecting the primary disciplines of the cited articles. The right side shows the journals where the cited literature appears, indicating the disciplines that frequently reference the material. Colored lines denote the citation connections between them.

In order to build on existing knowledge and concepts, researchers need to know which works have gained the most attention and citations. Table 3 presents the top 10 most cited articles in the field of artificial intelligence and Alzheimer’s disease research, with citation counts ranging from 103 to 321. The document with the highest global citations, “Alzheimer’s disease drug development pipeline: 2022” authored by Jeffrey Cummings and colleagues, appeared in the journal Alzheimers Dement (N Y) in 2022 (26). Following in second place is “Deep Learning in Alzheimer’s Disease: Diagnostic Classification and Prognostic Prediction Using Neuroimaging Data” by Taeho Jo and team, which was published in Frontiers in Aging Neuroscience in 2019 (27), garnering 281 citations. In third position, “Recent Advancements in Pathogenesis, Diagnostics and Treatment of Alzheimer’s Disease” by Sahil Khan and co-authors, earned 272 citations upon its publication in Current Neuropharmacology in 2020 (1).

A co-citation network is a graphical depiction of how frequently documents are cited together, indicating a conceptual link between them. Conceptual clusters form within this network when certain documents are consistently co-cited, signifying a shared research theme or focus. These clusters help identify central topics and influential works within an academic field. The top 10 co-cited references in the artificial intelligence and Alzheimer’s disease field are listed in Table 4. The most co-cited article is “The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease (28)” (55 citations), followed by “The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease (29)” (42 citations). The third most co-cited article is “Mini-mental state. A practical method for grading the cognitive state of patients for the clinician” and “NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease (30)” (38 citations). In order to construct a map of co-citations, references with a co-citation count of at least 15 were selected, a total of 52 references were selected for analysis of co-citations (Figure 6A). There were different colors for different clusters of references. The first cluster (in red) included 21 references, while the second cluster (in green) included 18 references. Cluster 3 (in blue) included 7 references and Cluster 4 (in yellow) included 6 references. Citation surge denotes a rapid escalation in the frequency of citations within a brief timeframe, indicating the focal points of research interest for a specific duration and tracking the evolution of these focal points across different time periods. This phenomenon is also utilized to assess the dynamics and trajectory of emerging research trends. The top 25 references with the strongest citation burst are shown in Figure 6B. Among these, the publication titled “NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease (30)” authored by Clifford R Jack Jr. and colleagues in 2018 in the journal Alzheimers Dement, exhibited the most pronounced citation burst (strength = 7.61), with the surge occurring between 2020 and 2024. This review proposes a biomarker-based definition of Alzheimer’s disease, viewing it as a continuous pathological process and providing researchers with a common language to explore the intreractions between pathological changes and cognitive symptoms. In addition, we utilized CiteSpace to visualize the co-citation network of references, spanning the years 2010 to 2024 with yearly intervals, highlighting the key reference nodes. As shown in Figure 6C, each dot represents a reference, and they were grouped into different clusters based on their intersecting research areas, “disease diagnosis” (#0) is the largest cluster, followed by “AI” (#1), “smart living” (#2) and “xai” (#3). Subsequently, the timeline of key clusters was charted using CiteSpace, as displayed in Figure 6D, a temporal analysis shows that the hotspots of research have shifted from the original #9 early prediction, #8 random forests, #6 microstructure and #2 smart living to #1 ai, #3 xai, #4 machine learning, and #0 disease diagnosis.

A total of 2,900 authors published related articles on artificial intelligence and Alzheimer’s disease based on the visualization chart of 530 screened publications. Figure 7A displays the top 10 authors ranked by publications. Among the authors, Castiglioni I and Liu Y are the leading contributors with the highest number of articles published, each having published 6 articles. Among the most local citation authors (Figure 7B), CASTIGLIONI I and SALVATORE C rank first, with a total of 28 local citations, respectively. Followed by JO T (25 local citations), NHO K (25 local citations) and SAYKIN AJ (25 local citations). We analyzed 24,305 co-cited authors, and those cited over 20 times were visually analyzed, as shown in Figure 7C, there were 87 authors who were cited more than 20 times in the analysis. Among them, 7 authors whose articles were cited more than 50 times, JACK CR stands out with the highest number of co-citations, followed closely by PETERSEN RC and SUK HI, which indicates their research is highly regarded and influential.

Keywords, as standardized and specialized vocabulary, provide a comprehensive reflection of the research subject matter. Analyzing frequently occurring keywords visually can facilitate a clearer understanding of the focal points and trends within the area of study. As a result of our analysis, we identified 2,544 keywords related to artificial intelligence and Alzheimer’s disease. Table 5 presents the top 20 most frequently occurring keywords along with their cumulative link strength. In the visualization, only keywords appearing at least 5 times were included, resulting in a total of 162 keywords that met the criteria and were displayed using VOSviewer. A similar colored node indicates that two keywords are within the same cluster, and the size of a node indicates the occurrence frequency of a keyword, whereas the distance between two words indicates the strength of a relationship between two words (Figures 8A,B). The most frequently used keywords were “Alzheimer’s disease,” “artificial intelligence,” “dementia,” “machine learning,” and “mild cognitive impairment” and so on (Table 5; Figures 8A,B).

Keywords that were more closely related were grouped into the same cluster and assigned the same color, thereby roughly indicating the research emphasis in recent studies. As show in Figure 8A, based on the results of this study, selected keywords can be roughly divided into 6 clusters, the cluster analysis revealed six distinct groups of keywords, each color-coded to reflect research themes. The red cluster, comprising 49 keywords, was predominantly associated with “Alzheimer's disease”, “artificial intelligence” and “machine learning brain”. The green cluster, containing 43 keywords, featured key terms like “deep learning”, “prediction” and “magnetic resonance imaging”. The blue cluster, with 29 keywords, concentrated on areas such as “dementia”, “biomarkers”, “risk” and “cognitive impairment”. The yellow cluster encompassed 17 keywords, highlighting “early detection”, “automated detection” and “artificial neural network”. The purple cluster, with its 12 main keywords, focused on “mild cognitive impairment”, “national institute” and “diagnostic guidelines”. Lastly, the light blue cluster primarily included “diagnosis”, “neuroimaging” and “PET”. Figure 8B presents an overlay visualization of author keywords, with blue indicating earlier appearances and yellow signifying more recent terms. In the past, keywords like “discovery” “image classification”, “association workgroups” and “computer-aided diagnosis” dominated the discussion, while recent topics include “early detection”, “networking”, “ChatGPT” and “AI”. In addition, CiteSpace was utilized to perform a log-likelihood ratio clustering analysis on the collected keywords, resulting in the identification of 14 distinct clusters, as depicted in Figure 8C. The most prominent clusters are Cluster #0, associated with deep learning, Cluster #1, focusing on neurodegenerative disease, Cluster #2, centered on sleep movements, and Cluster #3, concerning risk factors.