Abstract
Background
The field of urological tumor histopathology has long relied on subjective pathologist expertise, leading to diagnostic variability. Recent advances in digital pathology and artificial intelligence (AI) offer transformative potential by standardizing diagnoses, improving accuracy, and bridging healthcare disparities. This study conducted a 20-year bibliometric analysis to map global research trends and innovations in AI-driven urological pathology.
Methods
For this bibliometric analysis, literature from 2004 to 2024 was retrieved from the Web of Science Core Collection. CiteSpace, VOSviewer, and Microsoft Excel were used to visualize coauthorship, cocitation, and co-occurrence analyses of countries/regions, institutions, authors, references, and keywords in the field of AI for urological tumor histopathology.
Results
A total of 199 papers were included. Research on AI-driven urological tumor pathology has steadily increased since 2005, with a significant surge between 2020 and 2023. The United States made the largest contribution in terms of publications (131), citations (4725), and collaborations. The most productive institution was the University of Southern California, while Patel et al. and Epstein et al. were identified as the most active and most cocited authors, respectively. European Urology led in both publication volume and impact. Keyword analysis identified “machine learning,” “prostate cancer,” “deep learning,” and “diagnosis” as major research foci.
Conclusions
The integration of AI into urological tumor pathology demonstrates transformative potential, significantly enhancing diagnostic accuracy and efficiency through automated analysis of whole-slide imaging and Gleason grading, comparable to pathologist-level performance. However, clinical translation encounters critical challenges, including data bias, model interpretability (“black-box” limitations), and regulatory-ethical complexities. Future advancements hinge on developing explainable AI frameworks, multimodal systems integrating histopathology, radiomics, and genomics and establishing global collaborative networks to address resource disparities. Prioritizing standardized data protocols, fairness-aware algorithms, and dynamic regulatory guidelines will be essential to ensure equitable, reliable, and clinically actionable AI solutions, ultimately advancing precision oncology in urological malignancies.
Introduction
Urological tumor histopathology traditionally depends on pathologists’ expertise, leading to diagnostic variability due to differences in experience and resources—especially between developed and developing regions. 1 The lack of standardized criteria exacerbates inconsistencies, increasing misdiagnosis risks and underscoring the need for reliable, uniform methods. Digital pathology (DP) addresses this by digitizing slides, enabling remote consultations and standardized protocols. 2 Coupled with artificial intelligence (AI), DP automates image analysis, enhances tumor detection, and improves diagnostic reproducibility. 3 This integration represents a paradigm shift, potentially bridging healthcare gaps and ensuring equitable access to high-quality diagnostics.
Despite the rapid increase in publications, the research landscape of AI-driven urological tumor pathology remains fragmented, with significant gaps in understanding its evolution. Key challenges include shifts in methodological approaches, such as the transition from Convolutional Neural Networks (CNNs) to transformer architectures, and the influence of high-impact institutions on shaping the direction of research.4,5 Moreover, the translation of technical innovations into clinical practice remains slow and inconsistent. In light of these challenges, a systematic bibliometric analysis is essential to better comprehend the underlying knowledge structure of this field.
Understanding the dominant research themes—such as the varying focus on prostate cancer versus renal cell carcinoma—along with emerging trends like weakly supervised learning,6,7 will provide valuable insights. Additionally, evaluating the equity of research contributions, including geographic and institutional disparities, is crucial in addressing imbalances in the availability of datasets compared to clinical needs. 8 Identifying underserved areas, such as AI tools for bladder cancer, 9 and tracking progress against translational milestones will be vital in guiding future investments in this field.
This research conducts a bibliometric analysis of AI-driven urological tumor pathology from January 1, 2004, to July 29, 2024. By identifying key countries, institutions, and authors, and mapping the field's knowledge structure, the analysis reveals evolving research trends. This study presents a large-scale bibliometric analysis of AI-driven urological tumor pathology research from January 1, 2004, to July 29, 2024, using computational mapping techniques such as cocitation networks and keyword bursts to assess the field's evolution. By identifying key contributors, institutions, and emerging research trends, the analysis provides a roadmap for advancing clinically impactful AI solutions. The study aims to contribute to a better understanding of how AI innovations may shape the future of urological pathology, help identify core issues, support the acceleration of clinical translation, and address technological bottlenecks and ethical concerns, such as algorithmic bias and lack of interpretability.
Methods
Database and searching strategy
The Science Citation Index Expanded (SCI-Expanded, 1999–present) within the Web of Science Core Collection (WOSCC) by Clarivate Analytics was selected as the primary data source. This database is among the most extensive and commonly utilized in interdisciplinary research, housing a vast array of academic journals and publications, making it a dependable resource for bibliometric analysis. This study retrieved and analyzed relevant literature published between January 1, 2004, and July 29, 2024. To ensure the precision and reliability of the findings, two researchers independently performed the literature search. The search strategy was designed based on prior research and structured as follows: TS = (“artificial intelligence” OR “robotic” OR “expert system” OR “intelligent learning” OR “feature extraction” OR “feature mining” OR “feature learning” OR “machine learning” OR “feature selection” OR “unsupervised clustering” OR “image segmentation” OR “supervised learning” OR “semantic segmentation” OR “deep network” OR “Bayes network” OR “deep learning” OR “neural network” OR “neural learning” OR “neural nets model” OR “artificial neural network” OR “data mining” OR “graph mining” OR “data clustering” OR “big data” OR “knowledge graph”) AND TS = (“urological cancer” OR “renal cancer” OR “kidney cancer” OR “bladder cancer” OR “prostate cancer” OR “penile cancer” OR “urethral cancer” OR “testicular cancer” OR “urothelial cancer” OR “upper urinary tract carcinoma” OR “nephroblastoma” OR “Wilms tumor” OR “transitional cell carcinoma” OR “adrenal cancer” OR “adrenal gland cancer” OR “renal pelvis cancer” OR “germ cell tumor” OR “adrenocortical carcinoma” OR “pheochromocytoma” OR “collecting duct carcinoma” OR “medullary carcinoma of the kidney” OR “small cell carcinoma of the bladder” OR “squamous cell carcinoma of the bladder” OR “sarcoma of the bladder” OR “adenocarcinoma of the bladder”) AND TS = (“Pathology” OR “histopathology” OR “digital pathology” OR “whole slide imaging” OR “virtual microscopy” OR “digital microscopy” OR “digital slides” OR “virtual slides” OR “telepathology” OR “telemicroscopy” OR “computational pathology” OR “computer-aided pathology” OR “digital image analysis” OR “pathology images” OR “pathomics” OR “urine cytology” OR “histopathological images” OR “pathological images”) AND SU = (“Urology”) AND PY = (2004–2024) AND LA = (English). The search results were limited to original research articles and review papers. For each selected publication, metadata such as titles, authors, keywords, institutions, citations, journals, and references were stored in plain text format for subsequent analysis.
Data extraction and analysis
The retrieved documents were first imported into CiteSpace V (Version 5.8.R4, Drexel University, United States) for deduplication. After cleaning, the dataset was exported and further analyzed using Microsoft Excel 2023. Quantitative analyses were conducted to identify the most productive countries, journals, authors, institutions, and highly cited articles. The H-index, a widely accepted metric for evaluating the impact and productivity of researchers, was also extracted from the WOSCC for the selected publications.
Data visualization
Bibliometric analysis and data visualization were carried out using three main tools: CiteSpace V (Version 5.8.R4), VOSviewer (Version 1.6.20), and Microsoft Excel 2023. VOSviewer, a popular tool for bibliometric studies, offers three visualization modes: network, overlay, and density maps. In this study, VOSviewer was applied to analyze author–keyword co-occurrence, coauthorship among countries/regions, authors, and institutions, as well as cocitation patterns across journals and references. CiteSpace, developed by Chen et al., is another powerful tool for visualizing and analyzing bibliometric data. Compared with VOSviewer, CiteSpace is more focused on exploring the relationships and dynamics between different knowledge domains. It is especially valuable for monitoring the development of research frontiers, uncovering emerging trends, and forecasting potential future directions. In this research, CiteSpace was applied for cocitation analysis of authors and references, dual-map overlays of journals, and identifying citation bursts in keywords and references. Microsoft Excel 2023 was used to manage tabular data, organize analytical results, and create supplementary visualizations, enhancing both the clarity and interpretation of data trends.
Results
Trends in annual and cumulative publications on AI-driven urological tumor pathology research over the last two decades
The volume of research articles released over various time frames reflects the interest and growth patterns of a specific field. Based on our search methodology and selection process (as shown in Figure 1), we collected 199 articles pertaining to AI-driven urological oncology pathology from the WOSCC over the past 20 years. As illustrated in Figure 2, the quantity of research papers in this area has steadily increased since 2005. Notably, starting from 2008, there has been a significant rise in both the annual cumulative output and the number of newly published papers. This trend is particularly pronounced between 2020 and 2023, during which the annual publication count surged. This reflects the growing impact of AI on urological oncology diagnosis as technological advancements continue. It also indicates an increasing level of academic interest among professionals in the field. However, due to the data collection cutoff date of July 27, 2024, the cumulative and annual publication counts for 2024 show a slight decline.
Flowchart of the publication selection in the study.
Global trends in the number of annual and cumulative publications on AI-based urological tumor pathology research over the past 20 years.
Contributions of countries/regions
As shown in Figure 3(a), a country/region citation distribution map was generated using VOSviewer. The size of each node is proportional to the volume of publications from that country/region, and the thickness of the connecting lines represents the frequency of collaboration between countries/regions. The findings suggest that the United States acts as a key player in AI-based research in urological pathology. The thicker lines between nodes denote stronger collaboration, with the United States demonstrating extensive partnerships with several countries, including Germany, Italy, the United Kingdom, and China. Notably, Italy and Germany also exhibit closer collaboration within the European region. Figure 3(b) presents a geographic map of global collaborations. The size of each circle corresponds to the number of publications from each country/region, while different colors represent distinct collaboration clusters. The lines connecting nodes indicate academic partnerships between countries or regions. The collaboration network of the United States spans a broad range of countries, particularly in North America, Europe, and East Asia, which are key regions driving the advancement of AI technology in urological tumor pathology research. Table 1 shows the top 10 countries ranked by the number of publications in this field. The United States ranks first with 131 publications, nearly four times more than Italy, which ranks second with 31 publications. Furthermore, the United States ranks first in citation count, with 4725 citations, significantly exceeding all other countries and emphasizing its worldwide prominence in this field of study. Austria and Sweden, while having fewer articles (9 and 8 publications, respectively), achieved high average citation counts of 48.27 and 56.25, respectively, reflecting the high quality and impact of their research contributions.
(a) Country/region citation distribution map generated using VOSviewer. The size of each node is proportional to the number of publications from each country/region, while the thickness of the connecting lines indicates the frequency of collaboration between countries/regions. (b) Geographical distribution map of total publications by country/region, where different colors represent individual countries/regions.
Top 10 countries with the most publications.
Institutional collaborations and funding contributions
Figure 4(a) illustrates the institutional collaboration network in AI-driven urological tumor pathology research over the past two decades. The University of Southern California (USC) occupies a central position, signifying its leadership and substantial influence in this field. The connecting lines between nodes represent collaborations between institutions, with thicker lines indicating stronger and more frequent interactions. The Medical University of Vienna emerges as a key node in Europe, reflecting its significant research impact and strong transcontinental collaborations, notably with institutions such as USC and Monroe Dunaway Anderson Cancer Center in Texas. Meanwhile, AdventHealth Global Robotics Institute (AGRI), although a smaller node, has formed a closely knit cluster of collaboration with several Italian universities, such as Università Politecnica delle Marche and the University of Modena and Reggio Emilia, suggesting its important contributions to the field. Another notable institution, the Cleveland Clinic, stands out for its collaboration network, mainly through partnerships with other North American institutions, especially within the United States. The polar bar chart (Figure 4(b)) summarizes the number of publications, total link strength (TLS), and total citations of the top 21 high-output research institutions. All of these institutions are located in North America and Europe. Specifically, AGRI ranks first with 13 publications, followed by the USC and the Medical University of Vienna. In terms of TLS, USC, the University of Modena and Reggio Emilia, and Institut Mutualiste Montsouris rank as the top three, indicating strong institutional collaboration. In the field of AI-based urological tumor pathology, the DOT plot (Figure 4(c)) displays the top 10 most active funding institutions, with results showing that seven are based in the United States. Notably, the National Institutes of Health (NIH), the Department of Health and Human Services, and the NIH National Cancer Institute are among the most prestigious institutions globally. This dominance underscores the significant role of the United States in leading this research field, supported by its strong economic foundation and substantial research funding.
(a) Visualization graph of coauthorship relationships between different research institutions. The size of each node is proportional to the number of publications by each institution, and nodes of the same color indicate institutions with similar research directions or close collaborative relationships within the same cooperation cluster. (b) Polar bar chart illustrating the number of publications, total link strength (TLS), and total citations of the top 21 high-yield research institutions. (c) Ranking of the top 10 most active funding agencies in AI-based urological tumor pathology research.
Journal contributions and cocitation networks
According to the results presented in Table 2, European Urology published the highest number of papers, with 23 publications and 2462 citations, followed by World Journal of Urology and BJU International. Not only does European Urology lead in publication count but it also holds a strong academic influence with its high citation numbers, underscoring its impact in the field. Notably, while Journal of Urology had fewer publications (13 papers), its high h-index and relatively substantial citation count indicate that it maintains high standards for publication quality and substantial academic impact. As per the 2024 Journal Citation Reports (JCR), most of the top 10 journals fall within the Q1 category, with European Urology holding the highest impact factor (IF = 25.3). This journal's prominence, with its leadership in IF, publication volume, and citation frequency, solidifies its status as a premier journal in the field. Journal cocitation is an essential metric for evaluating academic impact and understanding a journal's role within a research network. As illustrated in Figure 5(a), the network visualization highlights core academic journals and their cocitation relationships in the application of AI within urological tumor pathology research. Each node's size represents the frequency of journal cocitations, while colors differentiate between distinct research topic clusters. The network is divided into five main clusters: the green cluster represents urology-related journals, the red cluster covers medical imaging and associated technologies, the blue cluster focuses on basic biology and oncology, the yellow cluster is associated with perspectives on oncology and pathology, and the purple cluster relates to pathology and radiology. The top three journals in terms of TLS within the green cluster are The Journal of Urology, European Urology, and BJU International. In the red cluster, Radiology, The Lancet Oncology, and European Urology dominate in terms of TLS. Meanwhile, in the blue cluster, the leading journals are Journal of Clinical Oncology, Prostate Cancer and Prostatic Diseases, and The New England Journal of Medicine. A dual-overlay map (Figure 5(b)) provides insight into knowledge flow and the interactions between different disciplines within the field. The results indicate that AI-based research in urological tumor pathology not only involves journals from the urology field but also spans literature from disciplines such as molecular biology, medical imaging, pathological diagnosis, and genetics. This demonstrates the interdisciplinary nature of this research area.
(a) Network visualization map of journal cocitation analysis generated using VOSviewer. (b) Dual-map overlay analysis of AI-based urological tumor pathology research journals generated using CiteSpace.
Top 10 journals related to AI-based research in urological tumor pathology.
Analysis of active authors and cocited authors
Table 3 presents the top 10 most productive and most frequently cocited authors in the field of AI-based urological tumor pathology research. The most productive authors are primarily associated with institutions in the United States and Europe. Among them, Patel et al., Gill et al., and Pascal et al. rank in the top three positions, contributing eight, seven, and seven papers, respectively. Although Pruthi RS and Wallen EM have published fewer papers, they each have accumulated 794 citations, demonstrating significant recognition and influence within the research community. The analysis of cocitations reveals that Epstein, Ficarra, and Menon are the top three most frequently cited authors, highlighting their central role and influence in this domain. As shown in Figure 6(a), the coauthorship network provides a detailed overview of the collaborative patterns among different authors in AI-based urological tumor pathology research. The clusters of nodes in various colors represent different research teams. The figure reveals that most researchers collaborate within their own teams, with only a few engaging in interdisciplinary collaborations, which facilitates academic exchange across different fields. As illustrated in Figure 6(b), the cocitation network demonstrates the associations between authors based on the frequency of cocitations. The results indicate that Epstein, Ficarra, and Menon are key contributors in AI-based urological tumor pathology research, highlighting their substantial impact and influence in the field.
(a) Author collaboration network map generated using VOSviewer. Clusters of nodes in different colors represent various research teams. (b) Author cocitation network diagram generated using CiteSpace.
Top 10 most productive authors and top 10 most cocited authors in AI-based urological tumor pathology research.
Analysis of references and cocited references
Table 4 lists the top 10 most cited papers in the field. The clinical research article by Jeff Nix 10 is the most cited, with 423 citations, followed by Yossepowitch et al. 11 and Patel et al., 12 with 302 and 291 citations, respectively. As shown in Figure 7(a), the results emphasize the relationships between documents in the field of urological tumor pathology research, with a focus on AI-based studies. Each node represents a reference, and the size of each node corresponds to its cocitation frequency. Nodes with higher cocitation frequencies are larger, indicating that these references are more frequently cited within the field. Several prominent nodes (highlighted in yellow) can be observed, with Nix standing out as a highly cocited document located at the central hotspot, indicating its pivotal role in the research field. Similarly, Yossepowitch et al., Goldenberg et al., 13 and Patel et al. have made significant contributions to the cocitation network. As illustrated in Figure 7(b), the results depict the temporal evolution of key research themes in AI-based urological tumor pathology, revealing development trends and emerging research hotspots. The five major themes identified are: “machine learning,” “biopsy,” “robotic surgery,” “bladder cancer,” and “urologic surgery.” This indicates that machine learning has recently become a research hotspot, particularly in applications related to urological pathology, demonstrating rapid growth. Since 2015, research on robotic surgery has experienced significant growth. In contrast, earlier research areas, such as urologic surgery, appear to have reached a relatively mature stage, with the research focus gradually shifting toward advanced surgical techniques. Figure 7(c) summarizes the top 25 references with the strongest citation bursts. The citation bursts in this field started in 2005, triggered by the influential publication by Ahlering. 14 The most recent citation burst began in 2022 and is still ongoing. Notably, the paper by Epstein et al., published in the American Journal of Surgical Pathology, 15 exhibits the strongest citation burst with an intensity of 4.19. The sustained high citation rates of these key references suggest that AI-based research in urological tumor pathology will continue to be a major focus in the coming years.
(a) Visualization of citation relationships between references using VOSviewer. The size of each node corresponds to its cocitation frequency. (b) Timeline of research hotspots in the application of artificial intelligence within the field of urological tumor pathology. (c) Visualization map of the top 25 most cited breakthrough references in AI-based pathology research for urological tumors.
Top 10 original articles on AI-based urological tumor pathology research.
Keywords co-occurrence analysis
As shown in Figure 8(a), we visualized the primary research themes and trends in the application of AI in urological tumor pathology over the past two decades. The size of each node is proportional to the frequency of occurrence of the corresponding keywords in the literature. High-frequency keywords such as “artificial intelligence,” “machine learning,” and “deep learning” indicate the critical role and extensive application of these AI technologies in urological tumor research, which aligns well with the central theme of this study. Notably, “prostate cancer” is positioned at the center of the graph as the largest node, signifying its prominence as a primary focus in urological tumor research. Other large nodes, such as “radical prostatectomy,” “biopsy,” “cancer,” “bladder cancer prognosis,” and “diagnosis,” represent key research topics in the field, covering a broad spectrum of aspects ranging from diagnostics and surgical interventions to treatment outcomes. This suggests that AI's application in urological oncology is not only extensive but also expected to continue driving the growth and development of this research field in the future. Therefore, AI-based pathology research in urological tumors is likely to remain a promising and persistent research hotspot.
(a) Visualization of keyword co-occurrence analysis using CiteSpace. (b) Visualization map of the top 25 keywords with the highest citation bursts in AI-based urological tumor pathology research.
Figure 8(b) presents the top 25 keywords with the strongest citation bursts over the past two decades. The earliest burst was associated with the keyword “experience,” which first appeared in 2005. Over time, the research focus has shifted to other keywords. For example, “artificial intelligence” displayed the highest burst intensity (6.22) from 2021 to 2024, highlighting a significant surge in research interest and citation frequency in this field in recent years. Similarly, keywords such as “deep learning,” “machine learning,” “prostate cancer,” “biopsy,” and “diagnosis” have shown strong citation bursts during the same period, and these bursts are still ongoing, further confirming the high research value of AI-based studies in urological tumor pathology. The latest citation bursts began in 2022 and continue to the present, with keywords such as “magnetic resonance imaging” and “system” emerging as new research focal points. This trend highlights the sustained and growing interest in integrating advanced imaging technologies and AI systems in urological tumor research.
Discussion
Bibliometric analysis provides quantitative insights into literature, surpassing systematic reviews by covering more ground and offering robust statistical analysis for visualizing research distribution and impact.16,17 With AI's rapid growth driving medical advancements, especially in automatic diagnosis, researchers struggle to track trends amid information overload. 18 This analysis helps visualize application fields and future trends, guiding researchers to explore key frontiers. 19
As shown in Figure 2, from 2004 to 2024, the number of publications on the application of AI in urological tumor pathology has steadily increased, with 53.8% of the relevant literature published in the last five years. This surge is attributed to the rapid development of AI technology in the medical field, particularly advances in deep residual networks 20 and spatially constrained CNNs. 21 The continuous optimization and application of these technologies, especially in handling complex pathological images, have significantly improved diagnostic accuracy and efficiency. 22 Additionally, recent advancements in hardware technology have provided a solid foundation for large-scale data processing and complex model training, further promoting the rapid adoption of AI in medical imaging. 23
As shown in Table 1, the United States ranks first with 131 articles, demonstrating its leading role in this field. Notably, scholars from Sweden lead in average citations per paper, with 56.25 citations. The United States not only leads in research output but has also established extensive collaborative networks with European countries such as Italy, Germany, and France. This international collaboration has propelled the application of AI in urological tumor pathology and driven technological advancements. Such collaborative efforts not only contribute to the rapid development of technology but also provide stable innovation support for this field. China, the only major developing country in the top 10, has published 12 related articles, but its average citations per article rank the lowest among the top 10 countries, indicating a significant gap in scientific impact compared to leading international nations. This suggests that China is still in a developmental stage in this field, and the international recognition and influence of its research outcomes need to be improved. This phenomenon highlights three key systemic constraints faced by developing countries: lack of technical resources, weak international collaboration, and funding shortages. To address these, we propose a dual strategy: creating data-sharing frameworks based on FAIR principles and establishing transnational AI pathology research consortia to enhance AI technology in urological tumor pathology. 24
As shown in Figure 3(c), the top 10 most active funding agencies in AI-based urological tumor pathology research are predominantly located in developed countries. Developing countries face various challenges in the application and development of AI technology, particularly in terms of limited research funding, access to high-level scientific resources, and advanced data infrastructure (37080740). International collaborative projects and joint research programs could help accelerate technology transfer and enable more developing countries to make breakthrough progress in AI-assisted urological tumor diagnosis.
North American institutions dominate AI-based urological tumor pathology research, with United States research institutions standing out in particular. As shown in Figure 4, eight of the top 10 institutions with the highest output are located in the United States, with the remaining two in Europe. These institutions have made significant contributions to advancing AI technology in urological tumor pathology. Among them, institutions such as AdventHealth's Global Robotics Institute have shown particularly notable research output, although their overall collaboration networks remain limited and have not fully expanded globally. This underdeveloped collaboration network may be related to their research focus being overly reliant on domestic projects and relatively fewer international collaboration opportunities. Data sharing could be a key factor in promoting global research collaboration, 25 as the development and training of AI models heavily depend on large-scale, high-quality clinical data. 26 By sharing diverse patient data and pathological images, the generalization ability of AI models and diagnostic accuracy could be enhanced. Moreover, this would not only promote the collective advancement of global medical technology but also contribute to a more equitable distribution of medical innovations.
As shown in Table 2, the top 10 journals related to this field are mostly classified as Q1 or Q2, with the journal European Urology publishing the majority of high-quality related research (Table 4). This indicates that AI-based urological tumor pathology research has gained widespread attention and plays a critical role in advancing scientific progress and medical practice. Researchers have introduced the latest AI algorithms, data analysis methods, and applications in urological tumor diagnosis through high-quality studies published in these journals, helping the medical community understand and adopt cutting-edge technological advancements. As shown in Table 3, five of the top 10 most prolific authors are from the United States, contributing over 50 papers collectively, with Patel et al. leading with eight articles. This demonstrates the crucial role Western countries, led by the United States, play in advancing AI technology in urological tumor pathology.
As shown in Figure 6, we present the author collaboration network map, highlighting three key researchers: Epstein et al., Menon et al., and Ficarra et al. Epstein, an authority in urological pathology, has made significant contributions to the diagnosis and grading of prostate cancer. His research, published in the American Journal of Surgical Pathology, forms the foundation for pathological diagnosis in urological tumors and is gradually being integrated into AI models for diagnostic assistance. 15 Menon, a pioneer in robot-assisted urological surgery, has developed multiple innovative techniques that, combined with AI, further enhance surgical precision and outcomes. 27 Ficarra, specializing in the surgical treatment of urogenital cancers, has published several highly cited papers in the field of urological oncology, with more than 200 citations, promoting the application of AI in both preoperative and postoperative pathological evaluations. 28 Strengthening collaborations with these expert teams will be crucial to advancing research in this field.
As shown in Figures 7 and 8, recent keyword frequencies are mainly concentrated on terms such as “machine learning,” “deep learning,” and “artificial intelligence,” further indicating the growing prominence of AI in tumor diagnosis. Artificial intelligence technology, especially machine learning and deep learning algorithms, has demonstrated enormous potential in the automatic analysis of complex pathological data and precise diagnosis. However, clinicians and pathologists often lack knowledge and understanding of AI-related technologies, leading to a lack of trust and reliance on AI models in routine diagnostics, which in turn hampers its widespread adoption. 29 As shown in Figure 8(a), prostate cancer is the dominant theme in the research landscape, with keywords such as “prostate cancer” and “bladder cancer” showing particularly high activity in the field of AI. In contrast, rare cancers, such as penile cancer, receive relatively less research attention in the AI domain. Meanwhile, keyword co-occurrence analysis and annual trend comparisons reveal that AI research in bladder cancer has shown more significant growth, while research on kidney cancer, although increasing, remains relatively sparse. The use of AI in rare urological tumors like penile cancer faces research gaps, primarily due to insufficient high-quality datasets, hindering AI model development. For example, penile squamous cell carcinoma had about 2070 new cases and 470 deaths in the U.S. in 2022, making it challenging to create diverse datasets for deep learning. 30 Even in well-studied tumors such as prostate cancer, AI faces unique challenges due to tissue heterogeneity, where tumor regions display varied cellular structures and genetic characteristics. These variations complicate tumor subtype classification, micro-metastasis detection, and treatment response prediction. To enhance AI accuracy and generalizability, models must be tailored to address these complexities, requiring localized approaches and more refined datasets.31,32
Future tumor diagnosis should no longer be limited to a single data source, such as histopathological analysis or radiological imaging. 33 Instead, multimodal integration technologies will become the core direction for diagnostic development. By integrating different data sources into a multimodal AI system, such as interpretable deep learning algorithms, proteomics, genomics, and radiomics, researchers and clinicians will gain a more comprehensive understanding of tumor diagnostics.34,35 The combination of multimodal data not only improves diagnostic accuracy but also provides a more precise basis for personalized treatment plans.36–38 Advances in AI are transforming the diagnosis and treatment of urological tumors, particularly in the automation of radiology and pathology. By digitizing whole-slide tissue images, AI can accurately analyze tumor structures and identify morphological phenotypes, thereby enhancing patient management. In Gleason grading and other pathological analyses, deep learning–based AI systems perform at a level comparable to pathologists. This progress improves diagnostic accuracy and efficiency, promising better clinical outcomes for urological tumor management.
The clinical application of AI models faces challenges, including data bias, model interpretability, and regulatory compliance. Artificial intelligence training datasets may be biased if they do not adequately represent the entire patient population (e.g., gender, age, race), leading to reduced diagnostic performance for certain demographic groups. Additionally, many AI models, particularly deep learning algorithms, are often referred to as “black boxes” due to their lack of interpretability, which hinders trust and acceptance among clinicians. Future research should focus on improving AI model interpretability to enhance clinical adoption. 39
As AI algorithms gain FDA approval, they introduce regulatory and ethical challenges, particularly concerning data sources, privacy protection, algorithm transparency, generalizability, and the update process as new data emerge. Liability in cases of prediction errors also needs consideration. These challenges can be mitigated by promoting transparency through open-sharing AI models, including source code, model weights, and metadata, with the research and medical communities. Additionally, the development of best practices for interpretability, fairness standards to address machine bias, and guidelines for continuous algorithm improvement is essential. While deployment costs and professional requirements remain concerns, these may decrease as issues of interpretability and fairness are addressed. 39
The integration of multimodal data, such as tissue pathology, radiology, and genetic information, is a key direction for AI applications in urological tumor pathology. However, developing effective multimodal AI systems faces several challenges, including data standardization, interoperability across data types, and the complexity of combining multiple data sources. Overcoming these barriers is crucial for improving diagnostic accuracy and enabling AI-driven personalized treatment decisions.23,25
Limitations
This study relied solely on the Web of Science database for literature retrieval, potentially omitting relevant studies from other databases such as PubMed, Scopus, and IEEE Xplore. Nonenglish publications were also excluded due to language restrictions, which may introduce geographical bias. Future research should consider integrating multiple databases for a more comprehensive dataset. Moreover, recent high-impact studies may be underrepresented due to citation data lag.
Conclusions
In conclusion, this study presents the first comprehensive bibliometric analysis of AI-assisted urological tumor pathology publications from 2004 to 2024. The results indicate that the application of AI in the diagnosis of urological tumor pathology is experiencing rapid growth. Developed countries, such as the United States, maintain a dominant position in this field, while developing nations like China are quickly catching up. Both research institutions and nations should further strengthen international and interdisciplinary collaboration, particularly in providing support to underdeveloped regions. Future AI research in urological tumor pathology should prioritize the development of interpretable AI models, creating multimodal systems that combine pathology, imaging, and genomics, and establishing global collaborative networks. Emphasizing data standardization, fairness algorithms, and dynamic regulatory frameworks will ensure the equitable, reliable, and effective use of AI in precision tumor diagnosis and treatment.
Footnotes
Acknowledgements
The authors extend our gratitude to the researchers and authors in the field of AI diagnostics whose published works have greatly contributed to this analysis. The authors also acknowledge the efforts of the software engineers who developed the visualization tools that enabled us to present our findings more effectively. Their dedication and contributions have been invaluable in the completion of this article.
Ethical considerations
Because this study did not involve humans or animals, and the underlying data were obtained from public databases, ethical approvals and informed consent were not required.
Author contributions
FD and YH were responsible for data collection and initial manuscript drafting. JD, KL, ZZ, and QL contributed to data analysis and visualization. YZ and JJ assisted with the literature review and interpretation of the results. ZZ and XQ contributed to the conception and design of the study and provided critical revisions to the manuscript. All authors reviewed and approved the final manuscript.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Guangdong Provincial Traditional Chinese Medicine Bureau Research Project Grant (Grant No. 20242004), the Science and Technology Planning Project of Guangzhou City (Grant Nos. 202201020582 and 2024A04J4159), the National Natural Science Foundation of China (Grant No. 82302304), the Doctoral Workstation Foundation of Guangdong Second Provincial General Hospital, China (Grant No. 2022BSGZ011), the Elevate Engineering Foundation of Guangdong Second Provincial General Hospital, China (Grant No. TJGC2022009), the Medical Scientific Research Foundation of Guangdong Province, China (Grant No. B2025376), and the Guangdong Provincial Clinical Research Center for Laboratory Medicine (Grant No. 2023B110008).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
All data used in this study are from public databases, as noted in the main text.