We conducted a literature search using the Web of Science Core Collection (WoSCC), known for its high-quality data and robust citation analysis tools. The search formula was: (TS = (((Ocular OR eye OR optic* OR “conjunctival” OR “conjunctivitis” OR “conjunctiv*” OR “ophthalm*” OR tear) AND biomarker*) OR “Ocular biomarker”)) AND TS = (Alzheimer* OR “Alzheimer disease” OR “Alzheimer’s disease” OR “Senile Dementia”; Costanzo et al., 2023; Kim et al., 2023). We restricted our search to English-language articles only. To minimize variations owing to database updates, we performed the literature retrieval on August 7, 2024. All relevant information, including publications and citation counts, titles, author details, institutions, countries/regions, keywords, and journals, was collected for further bibliometric analysis.

Microsoft Excel (version 2016), R-bibliometric (version 4.3.3), VOSviewer (version 1.6.20), and CiteSpace (version 6.3.R1) were used for data analysis and visualization. Excel was used for data collection, table generation, calculation of key indicators, and analysis of annual publication trends. R-bibliometric enabled the visualization of published article volumes and performance analysis(Donthu et al., 2021). The H-index, a crucial metric for evaluating academic contributions, was obtained from WoSCC to quantify the academic impact of countries/regions, institutions, and journals (Tahamtan et al., 2016). VOSviewer was employed to visualize collaborative networks among countries, institutions, and authors, along with the co-occurrence and coupling networks of journals and keyword co-occurrence networks. In these visualizations, node sizes represented the number of publications, while links indicated co-authorships, with their thickness reflecting the strength of collaboration. Different colors represented research clusters, and total link strength in collaboration networks measured the frequency of co-authorship, highlighting the extent of collaborative research (Liu et al., 2024). CiteSpace was utilized for keyword burst analyses, with time slicing configured to one-year intervals. The threshold for the top N keywords per segment was established at 5, and both the Pathfinder method and pruning of the merged network were employed (Sabe et al., 2022). This methodology produced a keyword timeline centered on ocular biomarkers and AD, accurately reflecting the specific parameter settings for each node.

A total of 623 eligible publications were analyzed, as illustrated in the data screening flowchart in Figure 1. These publications, sourced from 267 journals, span the years 1989 to 2024 and include contributions from 4,564 authors, five of whom are independent authors. On average, each publication featured 8.84 co-authors, with international collaborations accounting for 32.42% of the total authorship. The average age of the publications was 4.72 years, and they received an average of 28.05 citations. Our analysis identified 1,543 unique keywords. In 1989, the inaugural year of the publication period, only one article was published. Following a 15-year hiatus, the number of published articles began to increase steadily in 2005. Figure 2 illustrated a significant rise in publications, with the total increasing from 50 in 2019 to 74 in 2020, and peaking at 85 publications in 2023. A simple linear regression analysis produced the equation: y = 4.3985x − 19.067, with a coefficient of determination of 0.8526. This finding indicated promising growth in this research area.

As shown in Table 1, most of the published literature on ocular biomarkers and AD originated from the United States (137 publications), China (107 publications), and Italy (40 publications), collectively accounting for 45.59% of all studies. The United States led with a total of 716 publications and 3,990 citations, followed by China with 427 publications and 1,980 citations, underscoring their substantial influence in this research area (see Table 1 and Figure 3). Additionally, Sweden boasted the highest Proportion of Multiple Country Publications (MCP_Ratio) at 87.50, indicating strong collaborative efforts. Among the top 20 countries, Germany had the highest average citations at 62.0, closely followed by Turkey at 60.3, suggesting that Turkey produces high-quality research despite its limited international collaboration. Of the 61 countries and regions involved in international collaborations with at least one published article, the United States demonstrated the strongest total link strength (119), followed by the United Kingdom (112) and Switzerland (63; see Figure 4). The interconnected clusters illustrated in Figure 4 reflect a significant level of global collaboration within this field.

The top 10 institutions with the highest number of publications were illustrated in Figure 5. Harvard University and the University of London were tied for first place, each contributing 64 publications, followed by University College London with 55. The international collaboration network among 113 institutions that produced a minimum of four articles was depicted in Figure 6. Ten clusters were identified, indicating substantial connections among the institutions. The red cluster, representing institutions from the United Kingdom, was centered around University College London and the University of Oxford, with University College London exhibiting the highest number of collaborations (46). The institutions within the purple cluster, which included Sapienza University of Rome and the University of Perugia, are predominantly from Italy. The brown cluster, featuring the University of Western Australia and the University of Melbourne, was primarily composed of Australian institutions. The blue cluster, which mainly consisted of institutions from the United States, included Harvard Medical School and Boston University. The green cluster was largely comprised of Chinese institutions, such as Tongji University and Fudan University. This visualization suggested that institutions from the same country or region tend to aggregate within the same cluster.

A total of 4,564 authors contributed to publications in this field, with the top 20 most cited authors listed in Table 2. Verbraak Frank D. ranked first in both TP (9) and TC (53), underscoring his significant authority in this domain. Despite Ramirez Jose M. ranked second in TP (8), he was positioned 34^th in TC (30). Additionally, Figure 7 visualizes the co-authorship network of 150 authors with at least three articles. Ramirez Jose M. (50) and Salobrar-Garcia Elena (50) had the highest number of international collaborations, followed by De Hoz Rosa (43), Salazar Juan J. (43), and Verbraak Frank J (43). Three clusters were depicted, with the red cluster (containing Babiloni Claudio and Noce Giuseppe) being the largest. The green cluster was relatively small, containing Zetterberg Henrik and Xia Weiming, while the blue cluster was the smallest, with only two authors.

Details on the 20 most prolific journals, ranked by H-index, with the Journal of Alzheimer’s Disease taking the top spot, were provided in Table 3. These journals produced 248 publications, representing 39.81% of all retrieved studies. Biosensors and Bioelectronics had the highest Impact Factor (IF = 10.7), with Brain closely followed (IF = 35.5), highlighting their significant impact. Notably, most of these journals belonged to the Q1 and Q2 quartiles. Moreover, the Journal of Alzheimer’s Disease ranked first in TP (42) and third in TC (827), underscoring its influence in this field. Despite fewer TP (18), Investigative Ophthalmology and Visual Science (914), signifying the high quality of its publications. Figure 8a depicts the co-occurrence network of 92 journals with at least 2 occurrences, with the Journal of Alzheimers Disease (166) having the highest total link strength, followed by Plos One (115) and Alzheimer’s Research and Therapy (113), reflecting their central role in connecting with other publications. As for the coupling network analysis (Figure 8b), 92 journals were included with at least 2 couples. The Journal of Alzheimers Disease (10,956) ranked first, followed by Scientific Reports (6,951) and Frontiers in Aging Neuroscience (6,326), emphasizing their central role in addressing common research topics.

Keyword co-occurrence analysis explores the relationships between frequently paired keywords revealing connections across various topics and uncovering the underlying structure of knowledge within a field. From 623 articles 98 keywords with at least 8 occurrences were obtained. “Alzheimers-disease” had the highest total link strength (727) with 244 occurrences followed by “dementia” (578) and “mild cognitive impairment” (514) indicating their strong associations with other terms (Figure 9). More recent keywords such as “recommendations” and “definition” (highlighted in yellow) have begun replacing older terms like “national institute” and “mild cognitive impairment” (in green) and earlier terms like “in-vivo” and “binding” (in purple). This shift demonstrates the evolving research focus over the years. Six clusters were detected with cluster 1 (including 32 keywords) the biggest. Cluster 1 centered around “biomarkers” “biomarker” and “nanoparticles.” While cluster 2 (20) contained “mild cognitive impairment” and “national institute” with cluster 3 (17) followed closely including “Alzheimer’s-disease” and “dementia.” Moreover cluster 4 (11) cluster 5 (9) and cluster 6 (9) were relatively small comprised of “nerve” “definition” and “atrophy” separately.

The top 20 terms with the most significant citation bursts related to ocular biomarkers and AD from 2005 to 2024 were illustrated in Figure 10. The grey lines represented the overall timespan, while the red lines indicated the periods of citation bursts. The term “cerebrospinal fluid” (5.91) shows the largest citation burst, followed closely by “optical coherence tomography angiography” (4.78). In 2011, the terms “optic nerve” and “protein” marked the beginning of a new research trend, which has evolved into various topics in the years since. Since 2018, terms such as “thickness,” “csf biomarkers,” “neurodegeneration,” and “plaques,” and “optical coherence tomography angiography” have gained increasing attention and continue to be central to ongoing research, signaling promising advancements in the field (Table 4).

The investigation into ocular markers for the early asymptomatic detection of AD commenced in 1989. However, a notable increase in the published literature was not evident until 2012, with a more pronounced surge occurring in 2019. An analysis of the geographical distribution of this research indicates that the United States has emerged as the leading nation in the study of ocular markers and AD diagnostics, producing the highest volume of published studies and maintaining significant collaborative ties with other countries. Notably, the rate of international collaboration within the United States remains relatively low, suggesting a preference for domestic partnerships. Furthermore, a review of pertinent journals focused on ocular markers and AD reveals that researchers are increasingly specializing in fields such as neuroscience, ophthalmology, and gerontology, particularly in relation to AD. Prominent journals in this domain, including the Journal of Alzheimer’s Disease, Plos One, and Alzheimer’s Research and Therapy, reflect a growing emphasis on neuroscience and ophthalmology within this body of research.

Our findings reveal a strong internal collaboration network among organizations dedicated to studying ocular biomarkers for diagnosing AD. Rosa de Hoz, Elena Salobrar-García, Juan J. Salazar, Inés López-Cuenca, and José M. Ramírez are associated with Complutense University of Madrid in Spain, where they specialize in retina, ophthalmology, and vision science. Their most recent publication explored the potential relationship between changes in retinal layer thickness and AD, employing OCT to identify quantitative parameters linked to the disease (Sánchez-Puebla et al., 2024). In Italy, Claudio Del Percio, Giuseppe Noce, Claudio Babiloni, and Andrea Soricelli are affiliated with two institutions: Sapienza University of Rome and The University of Naples Parthenope, showcasing a strong co-authorship within the country. Their research centers on neurophysiology, brain function, cognitive neuroscience, and neurodegeneration. Their latest publication integrated imaging methods, such as PET-MRI and PET, to examine Resting-State Electroencephalographic Rhythms in Parkinson’s disease with cognitive impairment. Notably, Frank Verbraak has emerged as the leading author in this field, holding the highest number of publications, citations, and impact (H-index). His comprehensive review established that changes in retinal thickness may serve as indicators of neurodegeneration in AD, providing a theoretical foundation for the potential use of ocular markers in disease detection (den Haan et al., 2017).

Our keyword analysis highlights the evolving trends in ocular markers associated with AD research from 2005 to 2024. In the initial period (2005–2010), the emphasis was primarily on fundamental diagnostic techniques. Keywords such as “spectroscopy” and “cerebrospinal fluid” emerged in 2005 and 2006, respectively, underscoring a focus on traditional diagnostic methods and fluid biomarkers. During these years, the term “in vivo” gained popularity, indicating a growing adoption of live imaging techniques in studies involving living organisms. From 2011 to 2015, research expanded significantly, with terms such as “degeneration,” “optic nerve,” and “apolipoprotein E” reflecting a growing interest in the pathological mechanisms of AD as they relate to ocular structures. By 2015, the emergence of “oxidative stress” as a prominent topic signaled a shift toward exploring the molecular mechanisms underlying Alzheimer’s pathology. In the mid-2020s (2015–2020), the focus shifted to neurodegenerative processes. Keywords like “nerve fiber layer,” “multiple sclerosis,” and “dysfunction” indicated an increasing interest in the clinical implications of these processes for ocular health. Concurrently, terms such as “CSF biomarkers” and “thickness” highlighted advancements in understanding CSF components and retinal thickness as potential diagnostic tools. In recent years (2020 onward), keywords such as “neurodegeneration” and “optical coherence tomography angiography” have gained prominence, indicating that high-precision imaging techniques, particularly OCTA, are becoming essential for detecting ocular biomarkers. This shift reflects a growing emphasis on advanced neurodegenerative marker analysis. Over time, topics like “cerebrospinal fluid” and “in vivo,” which were initially prominent in early research, have gradually been supplanted by newer techniques such as OCTA and concepts like “neurodegeneration,” illustrating rapid advancements in diagnostic technology.

Recent studies have highlighted OCTA as a non-invasive ocular imaging technique that effectively assesses the peripheral vascular system of the retinal microcirculation (Curro et al., 2024; Sun et al., 2024; Vagiakis et al., 2024). A study utilized OCTA and PET to investigate temporal changes in physiological parameters among patients with preclinical AD. In the Aβ + cohort, vessel density (VD) demonstrated non-significant incremental changes over time in both the macular and optic nerve head (ONH) regions. Conversely, the Aβ− cohort exhibited significantly elevated VD in both the ONH and macular regions. Furthermore, the Aβ++ group revealed markedly greater VD in both the inner and outer rings of the macula when compared to the Aβ+ and Aβ− groups. These variations in VD appear to emerge early in the progression of preclinical AD, underscoring distinct differences between the ONH and macular regions (Curro et al., 2024). Although a consensus regarding the utility of the foveal avascular zone as a biomarker remains elusive (Curro et al., 2024; Vagiakis et al., 2024), OCTA continues to show promise in differentiating AD from other neurodegenerative disorders (Sun et al., 2024).

In addition to the analysis of retinal structure, emerging detection technologies have enabled the examination of tear components and their movement parameters, which may contribute to the diagnosis of AD. Tears may provide a valuable source of biomarkers for AD due to their easy collection and accessibility. Elongation initiation factor 4E (eIF4E) has been found exclusively in the tears of AD patients, suggesting its potential to differentiate between various neurodegenerative diseases (Kenny et al., 2019). Additionally, the overall microRNA content is elevated in the tear samples of individuals with AD, with microRNA-200b-5p emerging as a particularly promising biomarker among the various individual microRNAs (Kenny et al., 2019). In the context of functional markers, eye movements serve as sensitive, low-cost, and non-invasive indicators for assessing cognitive changes (Chaitanuwong et al., 2023). Research indicates that the anti-sweep task reveals the most pronounced abnormalities in both early-stage and advanced AD, which are linked to frontal lobe dysfunction resulting from the neurodegenerative processes associated with the disease (Chehrehnegar et al., 2022). Furthermore, individuals with both early-stage and advanced AD demonstrate a significant increase in eye hopping latency and frequency errors when compared to cognitively healthy individuals (Opwonya et al., 2022).

Furthermore, the development of artificial intelligence (AI) in AD diagnosis has gradually emerged as a significant area of interest. Deep Learning (DL), a subset of AI, has been increasingly applied for the early detection of AD, which relies heavily on imaging for accurate diagnosis(Lagun et al., 2011; Cheung et al., 2022; Sun et al., 2022; Zuo et al., 2024). Cheung et al. utilized a substantial dataset of 12,949 fundus photographs collected from a diverse global population. This dataset comprised 7,351 images from 3,240 healthy individuals and 5,598 images from 648 patients diagnosed with AD. The study achieved an accuracy rate of 86.3% in detecting AD within its bilateral internal validation dataset. In the test dataset, the accuracy for AD detection varied, ranging from 79.6 to 92.1% (Cheung et al., 2022). Li et al. (2024) found significant differences in eye movement patterns between healthy controls and those with AD or mild cognitive impairment (MCI), suggesting potential biomarkers for early detection. Further, eye-tracking metrics have shown efficacy in clinical settings for monitoring cognitive decline in the elderly(Li et al., 2024). Opwonya et al. (2023) demonstrated that eye-tracking tasks, such as prosaccade and antisaccade paradigms, could distinguish cognitively normal individuals from those with MCI based on latency and error saccades. Integrating eye-tracking into diagnostic frameworks alongside other biomarkers may enhance understanding of the progression risk in individuals with subtle cognitive impairments. Moreover, the study found that eye-tracking performance correlates with amyloid plaque burden, suggesting its potential as a biomarker for early neurodegenerative disease detection (Opwonya et al., 2023).

Recent studies modeling eye movement data have also employed deep learning techniques to identify eye movement patterns associated with AD during three-dimensional visuospatial memory tasks (Sun et al., 2022; Zuo et al., 2024). Ophthalmic AI exhibits improved performance when utilizing multimodal retinal imaging techniques, rather than relying solely on fundus images (Chaitanuwong et al., 2023). Despite challenges such as limited dataset availability, a lack of data diversity, and inconsistent image quality, deep learning, due to its exceptional predictive capabilities, is still at the forefront of biomarker research in the detection of AD.