We conducted a literature search on the Web of Science Core Collection (WoSCC) on November 12, 2024. The search queries were used to retrieve articles about fMRI and diabetes mellitus published between 1987 and 2024: Searched Method: TS = (functional magnetic resonance imaging* OR “functional MRI*” OR fMRI* OR (“functional” NEAR/3 “magnetic resonance”) OR ((“functional” NEAR/3 “imag*”) AND (magnetic OR resonance OR MR))) AND TS = (diabet* OR mellitus OR T1DM OR T2DM OR hyperglycemia OR glucose intolerance). All information was collected in plain text file format. The following basic information was collected for each article: titles, author information, institutions, countries/regions, keywords, journals, and references. Articles that met the following criteria were included: (1) published in English, and (2) articles on fMRI and diabetes mellitus, including original research. The exclusion criteria were (1) not related to fMRI in diabetes mellitus, (2) non-article document type (e.g., review, case report, letter, and conference abstracts), (3) duplicated publications, or (4) non-English language.

Relevant data were extracted from the retrieved literature bibliographies, and used Microsoft Excel to identify and calculate the bibliometric indicators. These indicators cover key aspects of publications, including annual publication counts, citation frequencies, average citation rates, journal names, journal impact factors, publication countries/regions, publishing institutions, and authors. Excel allows researchers to organize and analyze bibliometric data efficiently. During the visual analysis process, three powerful bibliometric analysis tools were used to analyze the academic data comprehensively. These tools are VOSviewer (version 1.6.20), CiteSpace (version 6.3.R1), and R 4.3.3.¹ VOSviewer is a versatile software tool that plays a crucial role in mapping institutional collaboration, author collaboration, co-authorship, citations, and co-citations (21). VOSviewer allowed the visualization and exploration of complex collaboration and relationship networks within the academic domain, gaining deeper insights into the interconnections among authors, institutions, and publications. VOSviewer was used for keyword co-occurrence analysis, and CiteSpace software was used for keyword burst detection to delve further into emerging trends and research hotspots within our study field. The study employed CiteSpace 6.1.R1 for keyword co-occurrence analysis, setting parameters for time slicing from January 1981 to November 2024 (the initial publications in this field appeared in 1981). The time slice was set to 1 year; node types: keywords. When the node type was keywords, the threshold (top N in each slice) was 5, and the pruning method was pathfinder + pruning merged networks. Based on the parameter settings of each node, visual analysis was conducted to generate a keyword timeline map for the field of “task-based functional magnetic resonance imaging” research. The size of the nodes represents the number of publications, the thickness of the lines represents the strength of the link, and the color of the nodes stands for different clusters or times. The H-index was employed to quantify the academic impact of individuals and journals, respectively. The H-index is a vital indicator for evaluating the academic contribution of researchers and could predict their future scientific achievements (22, 23). In this study, the H-index of each author was obtained from WoSCC. The G-index refers to the highest number of papers that receive H-index or more citations (24). The M-index, defined as (h-index)/(number of years since the author’s first published paper), characterizes the rise in the H-index over time.

In total, 706 eligible publications were analyzed in the present study. The flowchart of data screening is shown in Figure 1. The investigation showed that 4,793 authors contributed to producing 706 manuscripts in this study. These works were published in 311 journals from 1987 to 2024, citing 24,560 references (Figure 2A). Figure 2B presents the specific number of annual publications regarding the application of fMRI in diabetes mellitus. Since 2006, fMRI and diabetes mellitus research has shown an upward trend. The annual publication volume exceeds 10 articles. The greatest number of annual publications was 70 in 2021. In addition, the index function y = 1.71x-14.767 (R² = 0.8051) of the annual publication trend was determined to evaluate the changing trend between fMRI and diabetes mellitus studies.

Table 1 presents the top 20 countries/regions based on the number of publications. The United States of America (USA) ranks first (n = 931) regarding total publication (TP), followed by China (n = 756) and Germany (n = 270). The USA also has the leading position (n = 10,258) in total citations (TCs), followed by China (n = 3,299) and Netherlands (n = 2086). Besides, the majority of the articles are single-country publications (SCPs) (Figure 3A). For example, more than 180 articles are SCPs in China. In the USA, 130 articles are SCPs. Among the 26 countries involved in international collaborations with a minimum of three articles, the USA has the highest number of collaborations with other countries (n = 151), followed by the United Kingdom (n = 70) and Germany (n = 59) (Figure 3B).

Figure 4A reports the top ten institutions by article count and rank. Harvard University in the USA ranks first with 109 publications, and Eberhard Karls University of Tubingen in Germany ranks second with 76 publications. Figure 4B depicts the institutional collaboration network map. Among the 77 institutions involved in international collaborations with a minimum of five articles, Harvard University has the highest number of collaborations with other countries (total link strength = 46), followed by Beth Israel Deaconess Medical Center (total link strength = 42) and Brigham and Women’s Hospital (total link strength = 34).

Table 2 shows the basic information of the top 20 most productive journals in fMRI and diabetes mellitus. The journal with the highest impact factor (IF) was Diabetes Care (IF = 14.8), followed by Cardiovascular Diabetology (IF = 8.5). Diabetes has published the greatest TP of 26 publications, followed by Frontiers in Neuroscience (23 publications) and Diabetes Care (20 publications). In addition, 95% of the top 20 journals were categorized in the Q1 or Q2 Journal Citation Reports (JCR) region. Regarding citations, articles published in Neuroimage were cited most frequently (1,314 times), followed by Diabetes (1,053 times). Regarding H-index, Diabetes Care was in the leading position (H = 22).

Figure 5 depicts the network visualization map of the journal co-citation analysis. The co-occurrence networks of journals contained 65 journals with at least three occurrences. The three key journals with the highest total link strength in co-occurrence networks were Diabetes (total link strength = 254), Frontiers in Neuroscience (total link strength = 164), and Frontiers in Aging Neuroscience (total link strength = 109).

Table 3 presents the top 20 core authors. The top 20 authors published 210 articles collectively, totaling 9,499 citations. Regarding TP, SHAO YI leads with 16 publications, followed by QIU SHIJUN with 14 publications. Regarding TC, FRITSCHE ANDREAS leads with 962 citations, and PREISSL HUBERT follows with 910 citations. Regarding the H-index and G-index, FRITSCHE ANDREAS is in the leading position. Figure 6 displays the collaborations among authors. Among the 97 authors involved in international collaborations with a minimum of three articles, QIU SHIJUN has the highest number of collaborations with other countries (total link strength = 104), followed by TAN XIN (total link strength = 100) and LIANG YI (total link strength = 85). In the network visualization, the size of the nodes increases with the number of articles published by the authors. The larger the node, the more articles the author publishes.

A comprehensive keyword analysis of the selected articles was performed using “Author Keywords” from the Bibliophagy application and “Keywords Plus” provided by the VOSviewer application. In total, 116 keywords with a minimum of eight occurrences were identified (Figure 7). Table 4 displays the frequency distribution of the keywords in the top 20 most frequent occurrences. Among them, the top five keywords identified were “dementia,” “risk,” “brain,” “Alzheimer’s disease,” and “functional connectivity.”

In addition, the keyword burst detection was used to identify research hot spots. Figure 8 depicts the top 20 references with the strongest keyword bursts on research during the period of 1994–2024. The green line indicates the 1994–2024 timeframe, and the red line indicates the period during which the outbreak was sustained. Since 2017, the keywords “association,” “prevalence,” “default mode network,” “diabetic retinopathy,” “spontaneous brain activity,” “memory,” “dysfunction,” “impairment,” and “diagnosis” have been more prominently concentrated, indicating promising developments. Among the identified keyword bursts, those featured by the end of 2024 were “dysfunction,” “impairment,” and “diagnosis.”

The present study identified 706 articles about fMRI and diabetes mellitus published from 1987 to 2024. The overall publication numbers maintained a stable growth trend. The USA published the largest number of papers. Among the top ten institutions, the majority are from the USA, suggesting that institutions from the USA lead an important position in this field. Diabetes is the journal with the most publications. SHAO YI has the highest publication among all authors.

In terms of journals, Diabetes has the highest number of publications in fMRI and diabetes mellitus, followed by Frontiers in Neuroscience and Diabetes Care. A previous study published in Diabetes suggested that cognitive functioning among T1DM can be assessed with fMRI techniques (25). Other studies published in Frontiers in Neuroscience used fMRI to acquire brain change data (26). Among the top 20 most cited journals, Neuroimage has the most TCs. The journal Diabetes is ranked second in the number of papers published. Regarding the H-index, the journal Diabetes holds the leading position. In addition, Neuroimage and Diabetes are both from the USA, demonstrating that the USA provides an important platform for developing fMRI and diabetes mellitus.

Regarding publication volume, the most prolific author is SHAO YI from China. His research focuses on brain activity in patients with diabetic optic neuropathy (27, 28). QIU SHIJUN from China mainly studied the functional connectivity (FC) of patients with T2DM and mild cognitive impairment (29). However, regarding TCs, FRITSCHE ANDREAS from the University of Tübingen, Germany, holds the first position and has the highest H-index. He mainly assesses the visceral adipose mass MRI and their correlation with markers for insulin resistance and prediabetes (19). It is crucial to consider the number of publications and the quality of their articles to identify prolific authors.

The annual publications output in the field of fMRI and diabetes mellitus demonstrated a continuous and stable upward trend in the past 30 years. Regarding countries/regions, China and the USA are the leading contributors to fMRI and diabetes mellitus research. Although China ranks first in the number of publications, it has a low position in TCs, indicating a lower quality of articles. The USA ranks first in TCs. The most cited article from the US was published in the Proceedings of the National Academy of Sciences of the United States of America in 2001, with 1986 citations (30), indicating a higher quality of articles. In addition, seven institutions among the top 10 are from the USA. In other words, the USA is significant in the field of fMRI and diabetes mellitus. In the USA, 30.3 million (9.4%) adults had diabetes mellitus in 2017 (31). The country has a good foundation in medical research and provides significant financial support for research. Overall, with irreversible globalization, there is still a need for more collaboration between countries/institutions to promote research in this field.

Among the top 20 most frequent keywords, the top five keywords were “dementia,” “risk,” “brain,” “Alzheimer’s disease,” and “functional connectivity.” In addition, the top 20 most cited keywords, listed in Figure 8, reveal the potential hotspots of fMRI and diabetes research in the past 30 years. Firstly, the “risk,” “dysfunction,” “impairment,” and “dementia” keywords suggest that the current literature mainly focuses on diabetes mellitus and the risk of cognitive impairment/dysfunction. Increasing studies suggest that T1DM and T2DM are associated with decreased performance in multiple domains of cognitive function and structural and fMRI abnormalities of the brain (27, 32, 33). Therefore, some researchers began to establish dementia risk prediction modeling with fMRI data to predict cognitive impairment/dysfunction and dementia. For example, Samoilova et al. used MRI data to create a computer neural network model for predicting the development of cognitive impairment in DM based on brain neuroimaging techniques (34). There are few studies on the prediction of cognitive function in patients with T2DM based on fMRI (35). Therefore, future research should present a new solution idea, which extracts the characteristics from fMRI and establishes a cognitive function prediction model among diabetes mellitus patients.

Secondly, the “dysfunction,” “impairment,” and “diagnosis” keywords indicated that one of the research hotspots is to find potential biomarkers for early diagnosis of cognitive impairment/dysfunction. Diabetes mellitus is associated with cognitive decline and altered brain structure (32). FC can be used to reflect brain function, diagnose neurodegenerative diseases, and provide insights into pathophysiologic mechanisms (36, 37). Blood oxygen level-dependent (BOLD) signaling in fMRI reveals hemodynamic changes associated with neural activity and has been used to detect FC alterations in patients with T2DM and mild cognitive impairment (MCI) (38–40). Region-specific FC provides useful features for T2DM-MCI diagnosis. Hence, FC is a potential biomarker for assessing the degree of cognitive decline (41). In addition, the application of multiple resting-state fMRI (rs-MRI) indices can detect abnormal neural activities in different brain regions in patients with T2DM, and a decreased coupling trend between the amplitude of low-frequency fluctuation (ALFF) can help understand the early changes before cognitive impairment appears. Compared with other fMRI techniques, rs-fMRI is widely used because it requires the least number of patients, acquires signals more easily, and is proficient in recognizing functional areas in many different groups, even those with low cooperation. Therefore, rs-fMRI may be useful for detecting central nervous system impairment caused by T2DM with DPN (14). Many studies explored the fMRI imaging of the brain in T2DM cognitive dysfunction (42–44). Research found that hippocampal atrophy can be used as an early imaging marker for structural changes in brain regions of patients with T2DM cognitive impairment (45). These findings provide some new insights into the neural mechanisms of diabetes mellitus-related cognitive impairment.

This study is the first bibliometric analysis on fMRI and diabetes mellitus over the past 30 years, demonstrating a systematic overview of the field and providing guidance for future research. With the help of the present bibliometric analysis, researchers interested in fMRI and diabetes mellitus can get an overview of the field and quickly get up to speed on the latest research hotspots. However, there were some limitations to this study. Firstly, this study’s reliance on specific databases may have excluded relevant literature not indexed in the SCI-E database from WoSCC. Secondly, while bibliometric methods provide valuable insights, they have limitations in capturing the full scope and nuances of the research landscape. Finally, the generalizability of findings may be limited to the specific context of fMRI in diabetes mellitus research.