3.1 Global overview
A total of 2,358 documents were obtained for exploration in this study based on the search strategy and literature screening criteria. This comprises 2,117 articles and 241 review articles. The application of machine learning to the kidney, a relatively new area of research, has seen a linear increase in global publications since 2014. The mean number of citations per article was 13.06, with 36 documents exceeding 100 citations. Furthermore, the 2,358 papers utilized in this study were authored by 14,031 researchers from 91 countries and regions, 3734 academic institutions, and published in 799 journals. Additionally, these papers have been cited in 81,725 articles from 14,322 journals. The publication activities of medical literature in this field exert a considerable influence and engage a significant global audience, which is of great significance in promoting the research development of nephrology and fostering international cooperation.
3.2 Annual publications and citations
Figure 2 illustrates the temporal distribution of publications and citations for machine learning in renal research. In summary, there is a notable increase in the number of publications and citations in the field of literature about the integration of machine learning and kidney research. Machine learning, a subfield of artificial intelligence, is a widely utilized tool in various fields. However, its application in the renal field has been relatively limited. The annual publication volume and citation frequency were explicitly analyzed by extracting the corresponding specific values from Fig. 2 and presenting them in Table 2. The paucity of publications and citations in the literature related to the topic between 2013 and 2019 indicates the field's nascent exploration stage. However, the number of articles published increased significantly in the latter five years, with the highest number of articles published in that period being 661. At the same time, the number of citations for these articles reached 10,209 in 2023. So far, the number of articles published in 2024 has decreased, but still reaches 200. This indicates that digital information and artificial intelligence technologies are becoming a subject of increasing scrutiny by researchers in renal medicine in the context of big data. This area has become a new focus in the renal field.
Table 2
Annual publication and citation frequency table of relevant literature in the field of machine learning applied to renal medicine
| Years | Publications | Citations |
| 2013 | 0 | 1 |
| 2014 | 4 | 10 |
| 2015 | 15 | 20 |
| 2016 | 16 | 87 |
| 2017 | 27 | 136 |
| 2018 | 66 | 324 |
| 2019 | 132 | 916 |
| 2020 | 243 | 2579 |
| 2021 | 414 | 5152 |
| 2022 | 580 | 8066 |
| 2023 | 661 | 10219 |
| 2024 | 200 | 3411 |
3.3 Analysis of country/region
In order to ascertain which countries have made the most significant contributions to machine learning applied to the renal field, a study was conducted in which 91 countries and territories were analyzed. Table 3 presents the ten countries with the most publications in machine learning applied to the kidney. The table includes the number of publications, total citations, average citation frequency, and the degree of collaboration. This allows for examining and analyzing these countries and regions' scientific strength and international influence. A total of 79% of all papers published in the field were produced by researchers in these countries. Among the countries contributing research papers to the field, China stands out with the most significant number of articles, 721, representing 30% of the field's publications. Nevertheless, the number of citations and the average citation frequency of papers are lower than other countries. This indicates that there is still potential for enhancement in the quality of Chinese literature in this field. Researchers should, therefore, direct their attention to this matter. The United States was the next most prolific contributor, with 575 articles representing 24% of the total articles in this field. Despite having the second-highest articles, the United States ranked first in the total number of cited articles, average citation frequency, and national or regional collaborations. It can be reasonably concluded that the United States has a robust scientific research infrastructure and a significant investment in research. The U.S. is characterized by far-reaching academic influence and first-rate research in this field. A data review reveals that the United Kingdom and South Korea have a higher total number of citations and average citation frequency despite having fewer publications. This indicates some degree of academic influence in these two countries. In conclusion, the current application of machine learning in nephrology is a nascent field of exploration that is rapidly evolving. International collaboration is necessary to ensure development, representing a significant challenge that requires attention in the future.
Table 3
Top 10 most productive countries for machine Learning research in renal medicine
| Rank | Country | Articles | Citations | SCP | MCP | Freq | MCP-Ratio | Average articles citations |
| 1 | CHINA | 721 | 5640 | 624 | 97 | 0.306 | 0.135 | 7.8 |
| 2 | USA | 575 | 12186 | 426 | 149 | 0.244 | 0.259 | 21.2 |
| 3 | INDIA | 109 | 968 | 82 | 27 | 0.046 | 0.248 | 8.9 |
| 4 | KOREA | 92 | 1075 | 68 | 24 | 0.039 | 0.261 | 11.7 |
| 5 | JAPAN | 74 | 480 | 64 | 10 | 0.031 | 0.135 | 6.5 |
| 6 | UNITED KINGDOM | 73 | 1409 | 28 | 45 | 0.031 | 0.616 | 19.3 |
| 7 | ITALY | 61 | 819 | 32 | 29 | 0.026 | 0.475 | 13.4 |
| 8 | CANADA | 60 | 739 | 27 | 33 | 0.025 | 0.55 | 12.3 |
| 9 | GERMANY | 53 | 814 | 22 | 31 | 0.022 | 0.585 | 15.4 |
| 10 | FRANCE | 45 | 749 | 24 | 21 | 0.019 | 0.467 | 16.6 |
| Note: SCP, single-country publication; MCP, multi-country publication. |
Subsequently, a national or regional cooperation network map was created for countries with more than five articles (see Fig. 3a and Fig. 3b). The greater the size of the round nodes in the figure, the greater the number of articles issued. In Fig. 3a, the larger the round nodes, the more the number of articles issued; the line between the nodes represents the strength of the association; the thicker the line, the more the number of articles issued in cooperation between two countries or regions. As illustrated in the figure, a substantial amount of collaboration can be observed within this field in numerous countries and regions. The United States has the most extensive cooperative network, followed by China, the United Kingdom, and Canada. The United States and China have the most extensive cooperative relationship of any two countries, with Canada also demonstrating high collaboration. Furthermore, interdisciplinary collaboration between research institutes and scholars in different countries is essential to address the challenges in the renal field. Renal disease is a global health concern, and various countries have a vast repository of renal case data and research resources. Cross-country collaboration can facilitate data sharing and resource integration, enhancing research efficiency and the quality of outcomes. In light of these developments, China and the United States have responded by cooperating in academic exchanges and mutual learning to promote the complementation and enhancement of research results. The application of machine learning in the renal field is of great significance in improving diagnostic accuracy, predicting disease progression, and providing personalized treatment. China and the United States are engaged in collaborative efforts to advance scientific research and the clinical implementation of machine-learning technologies in the renal field.
Figure 3b presents the analysis of cooperation between countries and regions using the VOS viewer. The color of the nodes indicates the presence of different clusters, the size of the nodes reflects the number of publications associated with the country, and the node color represents the degree of inter-country cooperation. The cooperation between countries or regions is classified into seven clusters, each with a distinct color. The dark blue cluster primarily comprises the United States, China, Japan, Thailand, and other countries. The light blue cluster includes Canada and other nations. The red cluster encompasses the United Kingdom and other countries. The green cluster includes India, South Korea, and other countries. The dark yellow cluster includes Germany, France, Spain, and other countries. The orange cluster includes Italy, the Netherlands, and other countries. Finally, the purple cluster includes Poland, Austria, and Russia. The connectivity and colors between the nodes demonstrate a robust connection between these countries, suggesting that these countries are at the forefront of machine learning for kidney applications.
3.4 Analysis of the Institution
At present, a total of 3,734 organizations are engaged in the application of machine-learning methodologies for research purposes within the field of renal medicine. This paper uses CiteSpace's node size and centrality concepts to generate Table 4. The software threshold was set to 50, yielding 381 nodes, 1,904 connecting lines, and a density of 0.0263. The number of messages an organization sends depends on the node's size—consequently, the larger the node, the greater the number of messages sent. Centrality measures a node's importance, reflecting the institution's importance in the domain.[37] The institutions that published the most significant number of articles were Harvard University (88), the University of California System (88), and Harvard Medical School (65). Of the top ten institutions, seven are from the United States, while the remaining three are from France, China, and the United Kingdom. Harvard University's medical school was established in 1782 and has since become renowned for its world-class faculty and technical resources, which provide an excellent platform for machine learning applications in kidney research. Furthermore, the elevated centrality of the University of London (0.1) and Central South University (0.07) suggests that the two countries play a significant role in information transfer. This implies that these organizations have significant consequences when applying machine learning to kidney disease.
Table 4
Top 10 most productive organizations for machine learning research in renal medicine
| Rank | Institutions | Articles | Centrality |
| 1 | Harvard University | 88 | 0.1 |
| 2 | University of California System | 88 | 0.04 |
| 3 | Harvard Medical School | 65 | 0.01 |
| 4 | Mayo Clinic | 54 | 0.07 |
| 5 | Institut National de la Sante et de la Recherche Medicale (Inserm) | 51 | 0.05 |
| 6 | University of London | 46 | 0.1 |
| 7 | Central South University | 40 | 0.07 |
| 8 | Pennsylvania Commonwealth System of Higher Education (PCSHE) | 40 | 0.04 |
| 9 | University of Pennsylvania | 39 | 0 |
| 10 | US Department of Veterans Affairs | 38 | 0.09 |
In order to gain further insight into the nature of institutional collaboration in this field, we have plotted Fig. 4a and 4b for visual analysis. Figure 4a depicts the network of inter-institutional collaborations between 60 institutions that have published more than 15 papers. In terms of institutional collaboration, six collaborative networks are particularly noteworthy. Each network consists of at least two institutions, with three comprising ten affiliated institutions engaged in collaborative activities. In most cases, the organizations comprising a collaborative network are based in the same country. As illustrated in the accompanying figure, the red cluster comprises 22 institutions, all of which are located in China. This evidence indicates that the Chinese government is aware of the significance of science and technology innovation and academic research and is promoting inter-institutional collaboration through various policies and financial support to facilitate the development of machine learning.
For this study, Fig. 4b illustrates each journal's mean year of publication. The lower right-hand corner of this visualization displays a color bar, which indicates the mapping of the score bars to colors. Our analysis revealed that the average year of all journals was published after 2021. This further authenticates that machine learning applied to the renal field has a late start and still has significant room for exploration.
Concurrently, the display indicates that Chinese organizations have a relatively late start and a rapid development of research in this field compared to organizations in other countries. Notable examples include Sun Yat-sen University, Shanghai University, Chongqing Medical University, and Shanghai Jiaotong University.
3.5 Analysis of author and co-cited authors
It has been observed in the existing literature that authors' productivity can indicate the researcher's dedication to the field.[38] The most productive authors were identified by analyzing the authors of the 2,358 papers sent for inclusion. Furthermore, the authors analyzed to ascertain the scholars and core strengths of machine learning in this research area applied to kidney-related diseases. In order to gain insight into the nature of the collaboration between authors, we analyzed author co-citation. In bibliometrics, two authors cited in the same publication by a third author are considered co-cited authors. This interpretation indicates that the higher the co-citation frequency, the more closely the authors collaborate and the more significant the authors' influence.[39] Table 5 was generated by combining authors, author postings, co-cited authors, co-cited author postings, and total link strength. A higher total link strength indicates a more significant collaboration with other authors. The list of the top ten authors is based on the number of publications and co-citation frequency. As illustrated in the accompanying table, the three most prolific authors are Cheungpasitporn, Wisit (23 publications), Thongprayoon, Charat (23 publications), and Pattharanitima, Pattharawin (20 publications), with a total link lightness of 214, 214, and 204, respectively. The three most frequently cited authors were Breiman, I (255), Levey, AS (243), and Lundbery, SM (187), with a total link strength of 2,027, 2,091, and 1,699, respectively. This evidence indicates that these authors have substantially contributed more to the field. It is worth noting that having the most publications does not imply having the most co-citations. This phenomenon can be attributed to the fact that the scholars who have made the most significant contributions to a particular field or topic tend to be the authors with the highest number of co-citations. Their papers are of high quality and impact, which results in them being cited more often. Conversely, the authors with the greatest number of publications may be publishing in multiple fields or topics. Despite this, the impact of each paper may not be as high.
Table 5
Top 10 authors in terms of number of publications and co-citation frequency of machine learning research in renal medicine
| Rank | Authors | Articles | Total link strength | Co-authors | Co-citations | Total link strength |
| 1 | Cheungpasitporn, Wisit | 23 | 214 | Breiman, I | 255 | 2027 |
| 2 | Thongprayoon, Charat | 23 | 214 | Levey, AS | 243 | 2091 |
| 3 | Pattharanitima, Pattharawin | 20 | 204 | Lundbery, SM | 187 | 1699 |
| 4 | Mao, Michael A | 15 | 165 | Chen, TQ | 171 | 1483 |
| 5 | Kaewput, Wisit | 13 | 140 | Pedregosa, F | 141 | 934 |
| 6 | Cooper, Matthew | 12 | 126 | Kocak, B | 136 | 2390 |
| 7 | Leeaphorn, Napat | 12 | 126 | Collins, GS | 132 | 1353 |
| 8 | Tangpanithandee, Supawit | 10 | 109 | Thongprayoon, C | 119 | 1080 |
| 9 | Jadlowiec, Caroline C | 10 | 107 | Koyner, JI | 116 | 2023 |
| 10 | Krisanapan, Caroline C | 10 | 107 | Kellum, JA | 101 | 1472 |
Based on the principles established by the renowned scholar Price (1963), we designate the authors who have published at least two works in the field of study as the core authors of this research.[40] The VOSviewer software was employed to create visual representations of author collaboration and co-citation networks, as illustrated in Fig. 5a and 5b. We can count a total of 1886 core authors out of 14,396 authors in this field of publications combining machine learning and kidney, reflecting the increasing attention of international researchers to the application of machine learning in nephrology. Examples include diagnosis of chronic kidney disease and prediction of morbidity risk. Meanwhile, the graph of Rockhart's theorem was plotted using the visualization platform of R software, and the results of Fig. 5c show that the number of papers published by most authors is in the interval of 0–5 papers.
Figure 5a is plotted with the parameter of at least four publications per author, including 234 authors, to incorporate practical needs. Seven distinct clusters can be discerned, with thicker connecting lines indicating closer collaboration between the authors. It can be observed that authors within all three clusters engage in collaborative relationships and exhibit greater interconnectedness than in the other clusters. The map offers a comprehensive overview of the clusters to which authors belong, facilitating a more nuanced understanding of the collaboration and academic standing of the field.
The authors were filtered based on the number of co-citations, with those with less than 20 excluded. The resulting network was plotted, as shown in Fig. 5b. All co-cited authors were classified into five regions. The blue region included Bellomo, R., Bihorac, A., et al., the green region included Amer Diabet, Assoc, Bertsimas, D et al., the red region included Capitanio, U, Chen, X, et al., the yellow region included Thongprayoon, C, Van Buuren, S et al., and the purple region included Loupy, A, Reeve, J et al. It is also noteworthy that there are active relationships between co-cited authors within these groups and links between groups and related co-cited authors between groups. Notable examples include the following collaborations: Zhang, Y. and Breiman, L.; Kocak, B. and Breiman, L.; and others. Furthermore, authors from disparate countries collaborate, indicating that authors in this field prioritize the international sharing of resources and technology exchange.
3.6 Analysis of journal and co-cited journals
A bibliometrics platform and VOSviewer software were employed to identify the most prolific and influential journals and co-cited journals in machine learning applied to the kidney. The impact factor of the journal and the JCR partition introduced by Thomson Reuters can be used to assess the influence and academic status of the journal in this field, thereby enabling researchers to identify suitable journals for submitting their manuscripts.[41] Consequently, we present the impact factor and JCR partitioning of the top ten total journal publications and the top ten co-cited journal publications in the field, as shown in Table 6. A review of the literature published in this field over the past decade revealed that a small proportion of the articles were published in general interest journals, with the majority appearing in medical and medical information journals. A review of the top ten journals to which the literature belongs reveals that all of these journals contain more than 25 articles with an impact factor of more than 3. This is a considerable number, which fully reflects the scholarly contributions of these journals to the application of machine learning in renal medicine. The three journals with the highest number of publications were "Scientific Reports "(99), "Frontiers in Medicine" (41), and "Frontiers in Immunology" (40). Notably, except for "COMPUTERS IN BIOLOGY AND MEDICINE," the remaining journals are open-source publications. This indicates that the recent surge in the development of open-source journals has significantly contributed to advancing research in this field. Although some scholars have expressed reservations about the quality of open-source journals, the concept of sharing research results is widely recognized. A content analysis of the co-citation status of publications revealed that "PLoS One" and "Scientific Reports" were the most frequently co-cited publications, with a frequency of 1,641 and 1,628 times, respectively. The second most cited journal was Kidney International, with 1,597 citations. The second most cited journal was Kidney International, with 1,597 citations. Furthermore, it can be observed that three journals have an impact factor exceeding 100: "the New England Journal of Medicine" (IF = 158.5), "the Journal of the American Medical Association" (IF = 120.7), and "the Lancet" (IF = 168.9). Given the high impact factors and low co-citation frequencies among the ten journals in question, it can be reasonably assumed that articles published in "PLoS One" and "Scientific Reports" are more accessible. Notably, eight journals with the highest co-citation frequencies were classified as JCR region I. This indicates that these journals publish high-quality literature that is widely recognized in the field of nephrology. In summary, the factors above clearly indicate these journals' academic standing and influence.
Table 6
Top 10 journals in terms of number of publications and co-journals frequency of machine learning research in renal medicine
| Rank | Jounal | Articlies | IF | Q | Co-cited Journal | Co-citations | IF | Q |
| 1 | Scientific Reports | 99 | 4.6 | Q2 | PLoS One | 1641 | 3.7 | Q2 |
| 2 | Frontiers in Medicine | 41 | 3.9 | Q2 | Scientific Reports | 1628 | 4.6 | Q2 |
| 3 | Frontiers in Immunology | 40 | 7.3 | Q1 | KIDNEY INTERNATIONAL | 1597 | 19.6 | Q1 |
| 4 | PLoS One | 39 | 3.7 | Q2 | JOURNAL OF THE AMERICAN SOCIETY OF NEPHROLOGY | 1476 | 13.6 | Q1 |
| 5 | Journal of Clinical Medicine | 34 | 3.9 | Q2 | NEW ENGLAND JOURNAL OF MEDICINE | 1233 | 158.5 | Q1 |
| 6 | BMC Medical Informatics and Decision Making | 33 | 3.5 | Q3 | AMERICAN JOURNAL OF KIDNEY DISEASES | 1015 | 13.2 | Q1 |
| 7 | IEEE Access | 33 | 3.9 | Q2 | JAMA-JOURNAL OF THE AMERICAN MEDICAL ASSOCIATION | 949 | 120.7 | Q1 |
| 8 | COMPUTERS IN BIOLOGY AND MEDICINE | 29 | 7.7 | Q1 | Clinical Journal of the American Society of Nephrology | 898 | 9.8 | Q1 |
| 9 | Diagnostics | 29 | 3.6 | Q2 | LANCET | 861 | 168.9 | Q1 |
| 10 | Frontiers in Oncology | 28 | 4.7 | Q2 | NEPHROLOGY DIALYSIS TRANSPLANTATION | 834 | 6.1 | Q1 |
The scientific journals were arranged in descending order based on the number of papers published on the application of machine learning in renal disciplines according to Bradford's Law.[42] This resulted in forming a core zone dedicated to this field and several zones containing an equal number of papers from the core zone, as illustrated in Fig. 6a. The figure illustrates 27 core journals in the design field, with 728 publications representing approximately 30% of the total publications in the subject area. In order to more visually represent the collaboration between co-cited journals, a network diagram (Fig. 6b) was plotted. The figure divides the journals into five distinct clusters, each comprising at least ten journals. This suggests that the journals are engaged in active collaboration, forming more stable cooperative groups.
The biplot overlay is an analytical method that reflects changes in academic journals' disciplinary distribution, citation trajectories, and research centers.[43] The biplot visualization graphic (Fig. 6c) illustrates that the clusters on the left side represent the published journals of the retrieved literature. In contrast, the clusters on the right side represent the journals of the cited literature. In other words, the former represents cutting-edge research, the latter represents the basis of the literature, and the different colored trajectories in the graph represent the pathways through which the journals were cited. These pathways can be interpreted as representing the topics cited by the journals. As illustrated in Fig. 6c, there are four paths, each of a different color, primarily serving as references. The green path represents the application of machine learning to the field of kidney research in the journals of medicine, medical, clinical, neurology, sports, ophthalmology, dentistry, and dermatology. The green pathway represents research on the application of machine learning to the field of kidney disease in clinical, medicine-themed journals that primarily cite literature from cardiology, molecular biology, genetics, nursing, and medicine-themed journals. In contrast, the orange pathway indicates that literature from molecular biology and immunology-themed journals cites literature from molecular biology, genetics, and medicine-themed journals.
3.7 Keyword analysis
Keywords frequently mentioned in the same context can be identified through keyword co-citation analysis. This process enables the identification of the most active research areas and the emergence of new trends in machine learning applied to the kidneys. Three hundred thirty-two keywords with at least ten co-citation frequencies were identified using VOSviewer. The top ten keywords in the field were then listed, and Table 7 was constructed. The term "machine learning" was excluded from the analysis, and the most frequent term overall was "mortality," with 258 occurrences. This was followed closely by "artificial intelligence," with 255 occurrences. Furthermore, the total citation frequency and link strength of words related to prediction, acute kidney injury, classification, risk, outcomes, disease, and diagnosis are not low, indicating that these words are equally famous in this subject area.
Table 7
Top 10 co-cited keywords related to machine learning research in renal medicine.
| Rank | Keywords | Co-citations | Total link strength |
| 1 | machine learning | 1138 | 4969 |
| 2 | mortality | 258 | 1389 |
| 3 | artificial intelligence | 255 | 1301 |
| 4 | prediction | 241 | 1208 |
| 5 | acute kidney injury | 240 | 1268 |
| 6 | classification | 239 | 1069 |
| 7 | risk | 238 | 1174 |
| 8 | outcoms | 202 | 1096 |
| 9 | disease | 157 | 709 |
| 10 | diagnosis | 153 | 756 |
Subsequently, we conducted a keyword clustering analysis (Fig. 7a and Fig. 7b). The degree of interconnectivity between keywords is reflected in the thickness of the lines, while the magnitude indicates the frequency of occurrence. Figure 7a illustrates categorizing keywords into three clusters based on distinct research aspects. Cluster 1 is the red section of the document entitled "Management and Intervention in the Life Cycle of Patients with Kidney Disease." It primarily addresses "mortality," "acute kidney disease," "outcomes," "health," "prevalence," and "XGBoost," among other topics. Secondly, the green section is Cluster II, entitled "Advanced Technologies and Methods for Diagnosis, Monitoring, and Prediction of Kidney Diseases." This cluster is primarily concerned with the terms "machine learning," "artificial intelligence," "big data," "model," and "diagnosis." Finally, the blue part belongs to the third cluster, named "Exploration of biomarkers and therapeutic approaches for kidney diseases," which mainly includes the keywords of "biomarkers," "urine," "expressions," "immunotherapy," etc. The blue part belongs to the third cluster, "Exploration of biomarkers and therapeutic approaches for kidney diseases." To analyze the evolution of research trends within machine learning for kidney-related disease applications, we employed a color-mapping technique whereby keywords were color-coded according to the year they first appeared. (Fig. 7b) This approach allowed us to visualize the temporal evolution of research interest in this field. Lighter colors indicated that keywords appeared later, suggesting that they were more popular in recent years. Examples of such terms include "diabetic nephropathy," "immunotherapy," "signature," and "biomarker." The combination of these two perspectives and a specific nomenclature allows for the determination of the structural distribution of the field, thereby providing researchers with valuable information about the research and the various specific aspects between keywords and clusters.
Additionally, a keyword emergence analysis was conducted using CiteSpace to address the limitations of the graphs above in terms of changes in keyword saliency.[44] As illustrated in Fig. 7c, the 20 keywords that emerged between 2015 and 2024 and their respective intensities and temporalities are presented. As illustrated in the accompanying chart, the keywords that have experienced a surge in popularity over the past five years encompass a diverse range of topics, including "classification," "obesity," "pathology," "kidney transplantation," "diagnostic accuracy," "tumor heterogeneity," "angiomyolipoma," and others. The combination of the time of emergence, duration, and value of emergence of these burst words allows for the application of machine learning in the renal field to focus on the precision and individualization of medical treatment, with the help of technology to explore a specific type of complex disease.
3.8 Highly cited reference analysis
Co-citation reference analysis is a method used to assess the impact and importance of scientific papers or research results in the academic community. We performed co-citation reference analysis on 3258 references using CiteSpace software and then extracted relevant data. Based on the number of co-cited references, they were plotted in the top ten, and their relevant data were extracted, as shown in Table 8. With a co-citation frequency of 99 and a centrality of 0.03, the highest co-citation frequency was Tomasev N et al. 2018 in the journal "NATURE." The title of this article is "Clinically Applicable Methods for the Continuous Prediction of Future Acute Kidney Injury." The second and third most frequently cited papers were published in Critical Care Medicine by Koyner JL et al. (2018) and Tseng PY et al. (2020), respectively. Our analysis revealed that approximately half of the literature on machine learning techniques is related to acute kidney injury. This indicates that researchers are highly interested in applying machine learning to this aspect of acute kidney injury within the renal field. For example, "Machine Learning for the Prediction of Volume Responsiveness in Patients with Oliguric Acute Kidney Injury in Critical Care" and "Derivation and Validation of Machine Learning Approaches to Predict Acute Kidney Injury after Cardiac Surgery."
Table 8
Top 10 papers with the highest total citation frequency for references related to machine learning research in the field of renal medicine.
| Rank | Titles | Authors | Year | Co-citations | Centrality | Jounal | DOI |
| 1 | A clinically applicable approach to continuous prediction of future acute kidney injury | Tomasev N et al. | 2019 | 99 | 0.03 | NATURE | 10.1038/s41586-019-1390-1 |
| 2 | The Development of a Machine Learning Inpatient Acute Kidney Injury Prediction Model | Koyner JL et al. | 2018 | 66 | 0.01 | CRITICAL CARE MEDICINE | 10.1097/CCM.0000000000003123 |
| 3 | Prediction of the development of acute kidney injury following cardiac surgery by machine learning | Tseng PY et al. | 2020 | 55 | 0.01 | CRITICAL CARE | 10.1186/s13054-020-03179-9 |
| 4 | Machine learning for the prediction of volume responsiveness in patients with oliguric acute kidney injury in critical care | Zhang ZZ et al. | 2019 | 51 | 0.02 | CRITICAL CARE | 10.1186/s13054-019-2411-z |
| 5 | From local explanations to global understanding with explainable AI for trees | Lundberg SM et al. | 2020 | 44 | 0 | NATURE MACHINE INTELLIGENCE | 10.1038/s42256-019-0138-9 |
| 6 | Derivation and Validation of Machine Learning Approaches to Predict Acute Kidney Injury after Cardiac Surgery | Lee HC et al. | 2018 | 39 | 0.03 | JOURNAL OF CLINICAL MEDICINE | 10.3390/jcm7100322 |
| 7 | Global, regional, and national burden of chronic kidney disease, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017 | Bikbov B et al. | 2020 | 39 | 0 | LANCET | 10.1016/S0140-6736(20)30045-3 |
| 8 | Machine Learning in Medicine | Rajkomar A et al. | 2019 | 38 | 0.06 | NEW ENGLAND JOURNAL OF MEDICINE | 10.1056/NEJMra1814259 |
| 9 | Machine learning-based quantitative texture analysis of CT images of small renal masses: Differentiation of angiomyolipoma without visible fat from renal cell carcinoma | Feng ZC et al. | 2018 | 37 | 0.03 | EUROPEAN RADIOLOGY | 10.1007/s00330-017-5118-z |
| 10 | Deep Learning-Based Histopathologic Assessment of Kidney Tissue | Hermsen, M et al. | 2019 | 36 | 0.02 | JOURNAL OF THE AMERICAN SOCIETY OF NEPHROLOGY | 10.1681/ASN.2019020144 |
Figure 8a illustrates all co-cited references' cluster and dependency analysis, as visualized using CiteSpace software. As observed in the existing literature, the clustering is highly stable and persuasive when the Q-value is more significant than 0.3, and the S-value is more significant than 0.5.[45] By setting the parameters Q-value to 0.8441 and S-value to 0.9508, we identified 22 clusters. Of these, 11 demonstrated more stable groupings. Cluster 0 is the largest cluster and is concerned with the topic of "acute kidney injury." In a similar vein, Clusters 1, 2, and 3 are dedicated to the study of "renal cell carcinoma," "computational toxicology," and "chronic kidney disease," respectively. Additionally, several critical areas of interest are machine learning applied to kidney-related diseases. These include, but are not limited to, recent advances, future challenges, novel approaches, and renal fibrosis, among others. Furthermore, in this network graph, a node represents a piece of literature, thicker connecting lines represent the existence of a co-citation relationship with a piece of literature, and the colors of the different clusters represent the evolution of the study, with red bursts of varying sizes distributed in the clusters. The interconnectivity of disparate literary works across distinct color clusters can be utilized to ascertain the research knowledge base and the boundaries of current research. In recent years, medical researchers in this field have devoted considerable attention to several critical topics, including cluster 0, "acute kidney injury," cluster 1, "renal cell carcinoma," cluster 3, "chronic kidney disease," and cluster 4, "recent advance." These areas represent the most significant and cutting-edge developments in the field. The knowledge base is concentrated on three key areas: cluster 5, which addresses the challenge of future medicine; cluster 7, which concerns the issue of drug-induced nephrotoxicity; and cluster 9, which focuses on the use of data-driven clinical pathways.
Subsequently, a co-citation timeline graph analysis of the literature was conducted, as illustrated in Fig. 8b. Articles within the same cluster are considered equally important, with the number of cluster nodes indicating the relative significance of the cluster. Larger nodes indicate that the literature contributes more to the cluster. As illustrated in the accompanying figure, our analysis identified seven primary clusters. A high concentration of outbreaks can be observed in the co-cited references of Clusters 0, 1, 3, and 4. This observation is consistent with the previous section. In contrast, Clusters 6, 8, and 17 have received relatively little attention from researchers. The duration of clusters in the study area is variable, with some clusters exceeding ten years and others existing for a shorter period. For instance, the clustering of 0 and the clustering of 1 have been the subject of considerable research and debate, indicating that there are still significant challenges to overcome in these two fields.