Emerging trends in artificial intelligence research in lymphedema: An evaluation in light of bibliometric and altmetric data

Abstract

Background

This study aims to systematically evaluate the current landscape of artificial intelligence (AI) and machine learning applications in lymphedema research by employing bibliometric and altmetric analyses. The goal is to identify major trends, research focuses, and influential contributors in this rapidly evolving field.

Method

A total of 43 AI-related articles on lymphedema published between 1975 and 2025 were retrieved from the Web of Science Core Collection. Bibliometric indicators such as publication years, journals, countries, authorship, and citation metrics were analyzed. Altmetric scores were also assessed. Each study was classified by study type and thematic focus.

Result

Original research articles constituted the majority (n = 26), with clinical studies being the most common subtype. The United States and China led in publication output. Most studies were published in Q1-Q2 journals, indicating high scientific quality. Scientific Reports was the most productive journal. General AI applications and risk prediction emerged as dominant themes. A moderate positive correlation was found between average annual citations and altmetric scores (r = 0.470, p = 0.039), suggesting consistency between academic impact and online visibility.

Conclusion

This is the first study to comprehensively map AI-based research in the field of lymphedema using bibliometric and altmetric methods. The findings reveal increasing global interest and high-impact publications, particularly in the domains of risk prediction and early diagnosis. These insights may guide future methodological frameworks and interdisciplinary collaborations in this emerging field.

Keywords

lymphoedema

Introduction

Lymphedema is a chronic and progressive health problem characterized by the accumulation of protein-rich fluid in tissues as a result of dysfunction of the lymphatic system.^1,2 It is classified as primary (congenital abnormalities) or secondary (acquired damage).^3,4 One of the most common causes of secondary lymphedema is cancer treatment. In particular, axillary lymph node dissection and radiotherapy during breast cancer surgery are risk factors.^3–6 Lymphedema significantly impairs quality of life by causing physical symptoms such as pain, swelling, and limited mobility, as well as body image and psychosocial issues.^2,7–9 Additionally, it imposes a significant economic burden on healthcare systems and patients.

Early diagnosis and effective management of lymphedema are of critical importance in reducing long-term complications and morbidity.² Diagnosis involves limb circumference measurements, bioelectrical impedance spectroscopy, lymphoscintigraphy, and MR lymphangiography.^10–13 However, these methods can be time-consuming and costly. They may also be inadequate, especially in the subclinical stage when symptoms are not yet evident.^3,10

Recent evidence indicates that obesity-related lymphedema has become a leading cause of secondary lymphedema in industrialized countries, reflecting the growing global burden of metabolic disorders and sedentary lifestyles. Obesity-related lymphatic dysfunction is characterized by impaired lymphatic transport, chronic inflammation, and progressive tissue remodeling. These pathological changes complicate early recognition and timely intervention. In addition, emerging early detection tools, including bioimpedance spectroscopy, optical lymphatic scanners, and automated volumetric assessment systems, enable the identification of subclinical disease before irreversible structural changes occur. These technological advances highlight the increasing need for predictive and monitoring tools that integrate clinical, imaging, and digital biomarkers to support personalized lymphedema management.

Over the last decade, artificial intelligence (AI) and its subfields—machine learning (ML) and deep learning (DL)—have increasingly been integrated into healthcare research, with applications ranging from diagnostic support to treatment decision-making.¹⁴ Recent high-impact studies have demonstrated the transformative role of AI across healthcare domains, including medical imaging, predictive analytics, precision medicine, and clinical decision support, further highlighting its growing relevance in modern healthcare systems.^15–18 This technological progress has also attracted interest in lymphedema research. Notably, ML approaches have been explored for estimating the risk of lymphedema before clinical symptoms arise, based on accessible clinical and laboratory data such as blood parameters (e.g., complete blood count, serum tests) and treatment-related indicators (e.g., lymph node dissection, radiation dose).^3,14,19 In parallel, DL techniques have been applied to the evaluation of medical images in the context of lymphedema diagnosis and monitoring.^13,20 These emerging directions emphasize the importance of synthesizing current knowledge to better understand the scientific reach and thematic orientation of AI-related studies in this field.

Systematic analyses that trace publication trends and highlight the key contributors are essential for evaluating the progression of a research domain. Bibliometric analysis enables the quantitative assessment of scientific activity, including productivity patterns, collaboration networks, and thematic evolution, by considering metrics such as publication frequency, citation counts, institutional and country-level contributions.^19,21 Complementarily, altmetric analysis captures the social and digital footprint of scientific outputs by aggregating their dissemination across platforms like Twitter, Facebook, academic blogs, and news media, thus offering insights into their broader societal engagement.²¹

This study aims to explore the role of artificial intelligence and machine learning in lymphedema research by analyzing publications indexed in the Web of Science database from 1975 to 2025. Through bibliometric and altmetric evaluations, the study investigates scientific productivity, geographical and institutional representation, and prevailing research themes. The findings are expected to outline the current landscape and inform prospective research directions in the field.

Materials and methods

Study design and search strategy

Using the Web of Science Core Collection (Clarivate Analytics, Philadelphia, USA) database, a comprehensive literature search was conducted on 30 May 2025. Articles published between 1975 and 2025 were retrieved using the following search strategy: (“lymphedema” OR “primary lymphedema” OR “secondary lymphedema”) AND (“artificial intelligence” OR “machine learning” OR “deep learning” OR “neural network”). The complete search strategy is provided in Supplemental Material 1. The extended time frame (1975–2025) was intentionally selected to capture the historical evolution of AI-related research in lymphedema, demonstrate the long-term absence of relevant studies, and objectively document the recent surge in publications observed after 2022. Since the data used in the analysis were obtained from publicly available scientific publications, ethical approval was not required.

Article selection

Fifty publications related to the subject were identified based on the search results. The abstracts and/or full texts of these publications were reviewed independently by two researchers (NT, FB) and evaluated for relevance to the study scope. Disagreements during the evaluation process were resolved through consultation with a third researcher (MHT) to achieve consensus. Only articles addressing AI applications in lymphedema-related studies were included in the analysis. Seven articles deemed irrelevant were excluded, resulting in 43 articles eligible for detailed analysis. A PRISMA-style flow diagram summarizing the study selection process is presented in Figure 1.

Figure 1.

PRISMA-style flow diagram illustrating the literature screening and study selection process for studies included in the bibliometric and altmetric analysis.

Data extraction

Within the scope of bibliometric analysis, various parameters such as article title, year of publication, total number of authors, author names (first and corresponding author), number of citations, citation index, journal name, Q index, H-index, impact factor, authors’ countries, and article type were recorded. When authors were from different countries, the country of publication was determined based on the institution of the first author. The citation index was calculated by dividing the total number of citations of an article by the number of years since its publication. Articles were ranked from the most cited to the least cited; in the case of articles with the same number of citations, the citation index adjusted for years was taken into account.²¹

Altmetric data were collected to evaluate the impact of articles outside academic circles. This score is calculated automatically by software based on the interactions received by articles on social media platforms, news sites, blogs, and various online sources. Factors such as the number of mentions (volume), the type of platforms mentioned (source diversity), and the profile of the individuals or institutions sharing the content (quality) influence the determination of the score. Interactions from sources such as Twitter, Facebook, LinkedIn, blogs, and news sites contribute to the Altmetric score at different rates and highlight the interest in the studies beyond the academic world. These data were obtained from https://www.altmetric.com/.²¹

Statistical analysis and methods

Statistical analyses were performed using IBM SPSS Statistics v24.0 (Armonk, NY, USA) software. The Shapiro–Wilk test was applied to evaluate the distribution of the data. Within the scope of descriptive statistics, mean ± standard deviation and median (minimum–maximum) values were used for quantitative data, while frequency and percentage (%) values were used for qualitative data. Spearman’s rank correlation coefficient was used to evaluate the relationship between variables that did not show a normal distribution. A correlation coefficient of r ≥ 0.60 was considered strong, 0.30 ≤ r < 0.60 was considered moderate, and r < 0.30 was considered weak. The statistical significance level was set at p < 0.05 for all analyses.

Results

Annual growth trend of publications

Using the Web of Science Core Collection database, a search was conducted using the keywords “lymphedema” OR “primary lymphedema” OR “secondary lymphedema” AND (“artificial intelligence” OR “machine learning” OR “deep learning” OR “neural network”) for articles published between 1975 and 2025, resulting in a total of 50 publications.

After manual screening, 43 articles were deemed suitable for detailed analysis. The total number of citations for these 43 analyzed articles is 209. The number of citations per article ranges from 0 to 26. The average number of citations is 4.86. The median number of citations is 3.0.

The article with the highest number of citations is “Developing and validating a prediction model for lymphedema detection in breast cancer survivors,” published in 2021 by Wei et al., which received a total of 26 citations (Table 1). The article with the highest citation index is “Segmentation of Arm Ultrasound Images in Breast Cancer-Related Lymphedema” (Table 1). The total number of citations and the number of articles by year are presented in Figure 2. The average citation index score was found to be 1.39 ± 1.61, and the median citation index was 1.

Table 1.

General information related to the articles (sorted by total citations).

Rank	Title	Authors	NA	CA	Year	Journal	Study design	TC	CI	AS
1	Developing and validating a prediction model for lymphedema detection in breast cancer survivors	Wei, Xiaoxia; Lu, Qian; Jin, Sanli; Li, Fenglian; Zhao, Quanping; Cui, Ying; Jin, Shuai; Cao, Yiwei; Fu, Mei R.	9	Qian Lu	2021	European Journal of Oncology Nursing	· Cross-Sectional Study	26	5.2	1
2	Application of multiphoton imaging and machine learning to lymphedema tissue analysis	Kistenev, Yury, V; Nikolaev, Viktor V.; Kurochkina, Oksana S.; Borisov, Alexey, V; Vrazhnov, Denis A.; Sandykova, Ekaterina A.	6	Yury V. Kistenev	2019	Biomedical Optics Express	Original research article	22	3.1	26
3	Segmentation of Arm ultrasound images in breast Cancer-related lymphedema: A database and deep learning algorithm	Goudarzi, Sobhan; Whyte, Jesse; Boily, Mathieu; Towers, Anna; Kilgour, Robert D.; Rivaz, Hassan	6	Rivaz Hassan	2023	IEEE Transactions on Biomedical Engineering	Original research article	18	6	1
4	Swarm intelligence empowered three-stage ensemble deep learning for arm volume measurement in patients with lymphedema	Shokouhifar, Ali; Shokouhifar, Mohammad; Sabbaghian, Maryam; Soltanian-Zadeh, Hamid	5	Shokouhifar Ali	2023	Biomedical Signal Processing and Control	Original research article	14	4.6	0
5	Smart-Cuff: A wearable Bio-Sensing platform with activity-Sensitive information quality assessment for monitoring ankle edema	Fallahzadeh, Ramin; Pedram, Mahdi; Saeedi, Ramyar; Sadeghi, Bahman; Ong, Michael; Ghasemzadeh, Hassan	6	Hassan Ghasemzadeh	2015	2015 IEEE International Conference on Pervasive Computing and Communication Workshops (PERCOM Workshops)	Conference paper	13	1.1	3
6	Comparison of machine learning models for predicting the risk of breast cancer- related lymphedema in Chinese women	Wu, Xiumei; Guan, Qiongyao; Cheng, Andy S. K.; Guan, Changhe; Su, Yan; Jiang, Jingchi; Wang, Boran; Zeng, Linghui; Zeng, Yingchun	9	Yingchun Zeng	2022	Asia-Pacific Journal of Oncology Nursing	Retrospective cohort study	11	2.7	1
7	Quality of Life Predictors in patients with melanoma: A machine learning approach	Pinto, Monica; Marotta, Nicola; Caraco, Corrado; Simeone, Ester; Ammendolia, Antonio; de Sire, Alessandro	6	Pinto, Monica	2022	Frontiers in Oncology	Observational study	11	2.7	1
8	Risk prediction models for breast cancer-related lymphedema: A systematic review and meta-analysis	Shen, Aomei; Wei, Xiaoxia; Zhu, Fei; Sun, Mengying; Ke, Sangsang; Qiang, Wanmin; Lu, Qian	7	Qian Lu	2023	European Journal of Oncology Nursing	Meta-analysis	10	3.3	2
9	Knowledge discovery from a breast cancer database	Mani, S; Pazzani, MJ; West, J	3	‐	1997	Artificial Intelligence in Medicine	Conference paper	8	0.2	1
10	Algorithm-based risk identification in patients with breast cancer-related Lymphedema: A cross-sectional study	Nascimben, Mauro; Lippi, Lorenzo; de Sire, Alessandro; invernizzi, Marco; Rimondini, Lia	5	Nascimben, Mauro	2023	Cancers	Cross-sectional Study	7	2.3	2

NA; CA; TC; CI; AS.

Figure 2.

Annual distribution of publications (bars) and total citations (line) related to artificial intelligence and lymphedema between 1997 and 2025. The left vertical axis represents the number of publications, and the right vertical axis represents the number of citations.

Co-authorship and co-citation analysis for the authors are presented in Figure 3(a) (minimum number of articles: 1 and above) and Figure 3(b) (minimum number of citations: 3 and above). Among the 43 analyzed articles, the highest number of articles was published in 2024 (n = 17). All articles are in English.

Figure 3.

(a) Co-authorship network of authors involved in artificial intelligence–related lymphedema research (minimum number of publications = 1). (b) Author co-citation network based on references cited at least three times. Node size represents publication or citation frequency, and colors indicate collaboration clusters.

The study entitled “Segmentation of Arm Ultrasound Images in Breast Cancer-Related Lymphedema: A Database and DL Algorithm”, published in Computer Methods and Programs in Biomedicine in 2023, had a citation index score of 6. The article with the highest altmetric score was “Application of multiphoton imaging and ML to lymphedema tissue analysis” (AS = 26).

The author with the most publications is Lu Q, who has contributed to a total of 3 articles. He has received a total of 36 citations. According to the analysis of contributing institutions, Peking University is the institution with the most publications, with 3 articles.

When the journals with the highest number of publications among the analyzed 43 articles were evaluated, Scientific Reports ranked first with 4 articles (Table 2). It was followed by the European Journal of Oncology Nursing with 3 articles, and Frontiers in Oncology, Cancers, European Radiology, and IEEE Access, each contributing 2 articles. Among these, all journals except the European Journal of Oncology Nursing were classified in the Q1 or Q2 categories, and Scientific Reports had the highest H-index (347) and impact factor (3.8).

Table 2.

Journals of the articles (n ≥ 2).

Journal name	Number of articles	Impact factor	H-index	Q classification
Scientific Reports	4	3.8	347	Q1
European Journal of Oncology Nursing	3	2.7	74	Q3
Frontiers in Oncology	2	3.5	156	Q2
Cancers	2	4.5	157	Q1
European Radiology	2	4.7	182	Q1
IEEE Access	2	3.4	290	Q2

At the country level, the United States (n = 12, 27.9%) and China (n = 10, 23.3%) contributed the most publications. When examining the thematic distribution, studies focused on general AI applications and risk prediction stood out (Figure 4).

Figure 4.

Distribution of research themes in artificial intelligence–related lymphedema studies, showing the number of articles across major application areas, including general AI applications, risk prediction, diagnosis, imaging, segmentation, classification, and dataset development.

In this study, no statistically significant relationship was found between the total number of citations and the annual average number of citations, or between the total number of citations and the Altmetric score (p = 0.264, p = 0.156). In contrast, a moderate positive correlation was found between the annual average number of citations and the Altmetric score (Spearman’s r = 0.470; p = 0.039) (Figure 5).

Figure 5.

Scatter plot showing the relationship between average annual citation counts and Altmetric scores of the included articles, with the fitted regression line and 95% confidence interval.

Author collaboration network

Co-authorship network analysis identified a research collaboration structure consisting of 19 nodes and 30 links, reflecting the extent and intensity of scholarly interactions within the field (Figure 6). The average node degree was 3.16, indicating that each author collaborated with approximately three other researchers on average. This finding suggests an active but moderately interconnected research community.

Figure 6.

Co-authorship network of authors in artificial intelligence–related lymphedema research, illustrating collaborative relationships and cluster structures. Node size represents publication volume, and link thickness indicates collaboration strength.

Network density was calculated as 0.1754, demonstrating that only 17.54% of all possible collaborative links were present. Although this indicates the existence of established research partnerships, it also suggests considerable potential for further collaboration. The relatively low density reflects a partially fragmented structure, in which several research groups operate with limited interconnection.

Cluster analysis revealed five distinct collaborative groups. The largest cluster consisted of six authors, while the remaining clusters included three to four authors. These clusters likely represent research teams organized around shared thematic interests or joint projects. The presence of multiple clusters highlights both specialization and emerging interdisciplinary collaboration within the field.

Lu Q emerged as the most influential author within the network, characterized by the largest node size and the highest number of collaborative connections. This central position indicates a key role in knowledge dissemination and research coordination. Other notable contributors included Rogacki K and Gopalakrishnan M, who also demonstrated substantial connectivity and citation impact. These authors function as important intermediaries facilitating information exchange and collaborative activities.

Overall, the co-authorship network demonstrates an evolving collaborative structure with identifiable research hubs and moderate connectivity. While existing partnerships contribute to research productivity and visibility, strengthening inter-cluster collaborations may further enhance knowledge integration and scientific output in AI-related lymphedema research.

Institutional co-authorship network analysis

Institution-level co-authorship analysis identified a collaboration network consisting of 23 nodes and 29 links, representing academic institutions involved in AI-related lymphedema research (Figure 7). The overall network density was 0.1146, indicating a relatively sparse collaborative structure in which only a limited proportion of potential institutional connections were realized.

Figure 7.

Institutional co-authorship network in artificial intelligence–related lymphedema research, illustrating collaborative relationships and cluster structures among academic institutions. Node size represents institutional publication and citation impact, and link thickness indicates the strength of collaborative ties.

The low density suggests that inter-institutional collaboration in this field remains underdeveloped, despite the presence of several active research groups. While collaborative ties exist, substantial opportunities for broader institutional partnerships remain.

Cluster analysis revealed the presence of eight distinct collaboration groups. The largest cluster consisted of six institutions, whereas the remaining clusters contained fewer nodes, reflecting localized or project-specific collaborations. This clustering pattern indicates that institutional partnerships tend to form around specific research initiatives or thematic interests rather than through widespread network integration.

Peking University and Peking Univ emerged as the most influential institutions within the network, each exhibiting the highest citation impact and centrality. Their prominent positions reflect their leading roles in driving research productivity and facilitating collaborative activities. Other notable institutions included the Mayo Clinic and Northwestern University, which also contributed to strengthening inter-institutional linkages.

Overall, the institutional collaboration network demonstrates moderate connectivity with clear central hubs and multiple peripheral clusters. These findings suggest that strengthening partnerships between major research centers and less-connected institutions may enhance knowledge exchange, promote multidisciplinary collaboration, and improve the global integration of AI-driven lymphedema research.

International collaboration network analysis

Country-level co-authorship analysis revealed an international collaboration network comprising 20 countries connected by 20 collaborative links (Figure 8). The overall structure of the network was relatively sparse, reflecting limited cross-national research cooperation in the field of AI-related lymphedema studies.

Figure 8.

International collaboration network in artificial intelligence–related lymphedema research, showing co-authorship relationships and regional clustering among countries. Node size represents publication and citation impact, and link thickness reflects the strength of international collaboration.

The network density was calculated as 0.1053, indicating that only a small proportion of potential international collaborations had been realized. The average node degree was 2.00, suggesting that each country collaborated, on average, with two other countries. This finding highlights the presence of several isolated or weakly connected countries alongside a small number of more collaborative hubs.

Cluster analysis identified ten distinct groups within the network. One dominant cluster consisted of 11 countries, reflecting strong collaborative relationships among these nations, whereas the remaining clusters comprised single-country nodes. This pattern suggests that international cooperation remains concentrated within limited regional or thematic partnerships, while many countries continue to operate independently.

The United States emerged as the most influential country, exhibiting the highest node size and citation impact (94 citations), followed by China (80 citations). Other prominent contributors included Italy, South Korea, the United Kingdom, Russia, and Germany, which also demonstrated notable research productivity and network centrality. These countries functioned as key intermediaries facilitating knowledge exchange and collaborative activities.

Overall, the international collaboration network demonstrates moderate fragmentation with a small number of central hubs and multiple peripheral countries. Strengthening cross-border partnerships, particularly between highly productive and less-connected countries, may enhance research visibility, promote methodological standardization, and foster the global integration of AI-based lymphedema research.

Keyword co-occurrence analysis

Keyword analysis identified a total of 266 unique terms across the included publications. The thematic distribution of keywords indicated that the literature predominantly focused on cancer-related conditions and their diagnosis, management, and clinical outcomes. The predominance of author-defined keywords suggests that researchers tend to position their studies within well-defined conceptual frameworks, reflecting thematic clustering and consolidation within the field.

The most frequently occurring keyword was “lymphedema” (n = 17), confirming its central role in current research. “Breast cancer” ranked second (n = 10), highlighting the close association between cancer treatment and secondary lymphedema. Keywords related to artificial intelligence, including “machine learning” (n = 10) and “deep learning” (n = 4), further demonstrated the increasing integration of computational approaches into clinical research.

Overall, the keyword network revealed the coexistence of clinically oriented studies and technologically driven analytical approaches, indicating a progressive shift toward data-informed lymphedema management (Figure 9).

Figure 9.

Keyword co-occurrence network and frequency distribution in artificial intelligence–related lymphedema research. The bar chart shows the most frequently used keywords, the line graph illustrates their temporal trends, and the word cloud visualizes overall thematic prominence. Larger fonts and node sizes indicate higher keyword frequency.

Thematic evolution analysis

Thematic evolution analysis was conducted to examine temporal changes in research themes between 1997 and 2025 (Figure 10). The analysis included 43 publications from 291 authors across 34 journals and identified a total of 266 unique keywords, indicating a dynamic and evolving research landscape.

Figure 10.

Thematic evolution of major research topics in artificial intelligence–related lymphedema literature between 1997 and 2025. The figure illustrates temporal changes in dominant themes, highlighting the recent growth of breast cancer– and surgery-related research and the relative decline of general diagnostic and machine learning–focused topics.

In the early period (1997–2011), research activity was limited and primarily focused on descriptive and clinically oriented topics. From 2012 to 2016, an initial increase in publications was observed, accompanied by emerging interest in diagnostic and imaging-related themes. A marked expansion occurred after 2017, reflecting growing academic attention to lymphedema and its technological assessment.

In the most recent period (2022–2025), “breast cancer” and “surgery” emerged as dominant and rapidly growing themes, indicating increased emphasis on cancer-related interventions and postoperative outcomes. In contrast, themes such as “lymphedema,” “machine learning,” “diagnosis,” and “arm lymphedema” showed relatively declining trends, suggesting a gradual shift toward more specialized and procedure-focused research directions.

Overall, the thematic evolution pattern demonstrates a transition from general disease characterization toward more targeted clinical and surgical applications. These findings highlight the increasing integration of technological and interventional perspectives in contemporary lymphedema research.

Supplementary qualitative and methodological analysis

To complement the quantitative bibliometric findings, a qualitative overview of representative and high-impact studies was conducted to summarize the principal AI methodologies, validation strategies, clinical applications, and key findings identified in the literature (Table 3). Additionally, a methodological classification of the included studies was performed to summarize study designs, AI methodologies, and validation strategies across the analyzed literature (Table 4).

Table 3.

Qualitative overview of representative and high-impact studies investigating artificial intelligence applications in lymphedema research.

Author	Year	Study design	AI method	Validation method	Clinical application	Key findings
Wei et al.	2021	Retrospective observational	Machine learning prediction model	Internal validation	Risk prediction	Developed predictive model for lymphedema detection in breast cancer survivors with favorable discrimination.
Kistenev et al.	2019	Experimental/technical development	Machine learning	Internal validation	Tissue analysis	AI-assisted multiphoton imaging improved tissue characterization in lymphedema.
Goudarzi et al.	2023	Technical development	Deep Learning	Internal validation	Ultrasound Segmentation	Deep learning algorithm demonstrated high performance in arm ultrasound segmentation.
Fu et al.	2018	Observational	Machine learning	Internal validation	Detection	ML identified lymphedema among breast cancer survivors with promising accuracy.
Du et al.	2024	Cross-sectional	Comparative ML models	Cross-validation	Risk prediction	Multiple ML models showed moderate-to-high predictive performance.
Lewsirirat et al.	2025	Technical development	Deep learning	Internal validation	Diagnostic classification	DL-based classification accurately differentiated lymphedema from other edema disorders.
Na et al.	2025	Technical development	Deep learning	Internal validation	CT evaluation	Automated CT segmentation improved lower-extremity tissue assessment.
Wang et al.	2024	Diagnostic Observational	AI-assisted ultrasound	Internal validation	Metastasis prediction	AI improved ultrasound-based prediction performance.

Table 4.

Methodological and technical classification of included studies investigating artificial intelligence applications in lymphedema research.

Variable	Category	Frequency (n)	Percentage (%)
Study design	Retrospective observational	8	18.6
	Prospective observational	4	9.3
	Cross-sectional	4	9.3
	Experimental/technical development	20	46.5
	Review articles	7	16.3
AI model type	Machine learning	16	37.2
	Deep learning	14	32.6
	Neural network	4	9.3
	Hybrid/ensemble models	9	20.9
Validation method	Internal validation	18	41.9
	Cross-validation	10	23.3
	External validation	5	11.6
	Not reported	10	23.3

Discussion

The global rise in life expectancy and improvements in cancer therapies have brought lymphedema and its associated comorbidities into sharper clinical and research focus.^22–24 In this evolving context, examining the literature on artificial intelligence in lymphedema research provides valuable insight into current directions and future developments.^20,25,26

Bibliometric and altmetric analyses provide complementary tools for evaluating scientific productivity, thematic trends, collaboration networks, and digital engagement beyond academia.^27–29 Although prior bibliometric studies have examined general or breast cancer-related lymphedema, no dedicated analysis has specifically focused on publications investigating artificial intelligence applications in this field. To the best of our knowledge, this study is the first bibliometric and altmetric analysis to systematically examine this specific intersection.

Article citations remain important indicators of scientific impact and visibility, although they are influenced by factors such as journal prestige, topical relevance, and publication time.³⁰ In the present study, the most cited articles primarily focused on AI-based detection, risk prediction, and medical image analysis, highlighting the core methodological priorities of this emerging field.^20,31–33

The 2023 study on ultrasound image segmentation in breast cancer-related lymphedema achieved the highest Altmetric score, reflecting the growing digital visibility of AI-driven research in this field.³² This finding further suggests that artificial intelligence research may enhance public and professional engagement with clinically underrepresented conditions such as lymphedema.^8,26,34,35

The United States emerged as the most prolific contributor, consistent with previous bibliometric reports.¹ The widespread availability of large-scale health datasets and interdisciplinary research infrastructure has facilitated the development of AI-based predictive models, particularly in cancer-related lymphedema. Similarly, Peking University demonstrated notable productivity, reflecting China’s strategic investment in healthcare-related artificial intelligence. These findings highlight the importance of institutional support and research infrastructure in shaping scientific output.

The relatively low institutional and international collaboration density observed in this study suggests a fragmented research landscape in AI-related lymphedema research. Limited collaboration may hinder multicenter validation, reduce the external generalizability of proposed models, and slow the dissemination of methodological innovations across institutions and countries. Strengthening international consortia, multicenter registries, interdisciplinary partnerships, and open-data initiatives may enhance collaborative efficiency, promote methodological standardization, and accelerate scientific advancement in this emerging field.

The diversity of journals publishing AI-related lymphedema research illustrates the interdisciplinary nature of this field, spanning oncology, radiology, nursing sciences, biomedical engineering, and data science.^16,26,27 This pattern underscores the importance of cross-disciplinary collaboration in addressing complex chronic conditions such as lymphedema.^3,26,36–38 The predominance of publications in Q1–Q2 journals further reflects the growing academic recognition and methodological rigor of research in this area.^26,39,40

The marked concentration of publications between 2022 and 2025 indicates a rapid expansion phase corresponding with broader advances in medical artificial intelligence.^22,39 This surge suggests a paradigm shift toward data-driven and algorithm-supported clinical frameworks in lymphedema research. However, as most studies remain focused on general AI applications, risk prediction, diagnosis, and imaging—with relatively limited exploration of advanced algorithmic areas such as segmentation and dataset development—the field still appears to be in an early translational stage with increasing emphasis on clinical applicability.

The observed positive correlation between annual citation averages and Altmetric scores indicates that early digital visibility may accelerate academic dissemination.⁴¹ However, both citation-based and Altmetric metrics should be interpreted cautiously. Citation-based measures may disadvantage recently published studies because of citation lag despite their potential scientific relevance. Likewise, Altmetric Attention Scores should not be interpreted as direct indicators of scientific quality, methodological rigor, or clinical validity, as they primarily reflect online visibility, media dissemination, and public engagement. Although Altmetric indicators provide valuable insight into digital and societal attention, they may reflect dissemination patterns and social interest rather than direct scholarly impact. Together, these findings suggest that citation and Altmetric metrics should be viewed as complementary—but distinct—indicators of academic and societal influence.

Clinical implications of artificial intelligence in lymphedema management

These clinical implications reflect the dominant themes identified in the bibliometric and thematic analyses, particularly diagnosis, risk prediction, and imaging-based assessment. Recent advances in artificial intelligence offer significant opportunities to improve lymphedema management through earlier diagnosis, risk stratification, long-term monitoring, and cost-effective care delivery. Given the increasing prevalence of obesity-related lymphedema and the complexity of its underlying mechanisms, AI-driven models integrating anthropometric, metabolic, and clinical risk factors may facilitate earlier identification of high-risk individuals and support preventive strategies.

Early detection remains fundamental to effective lymphedema management. Artificial intelligence-based image analysis, bioimpedance interpretation, and automated ultrasound segmentation systems have shown promising potential in detecting subtle tissue alterations during subclinical stages. These technologies may complement conventional diagnostic approaches by improving sensitivity, reducing operator dependency, and promoting standardized assessments across clinical settings.

Risk prediction represents another critical application area. ML algorithms trained on demographic, laboratory, treatment-related, and imaging data have demonstrated increasing accuracy in predicting lymphedema development, particularly among cancer survivors. Such tools may support personalized surveillance strategies, optimize follow-up schedules, and enhance preventive care planning.

Long-term monitoring is essential due to the chronic and progressive nature of lymphedema. Wearable sensors, mobile health platforms, and AI-assisted image analysis systems enable continuous and remote assessment of limb volume, tissue composition, and symptom progression. These technologies may improve patient adherence, facilitate early detection of disease exacerbations, and enhance access to specialized care, especially in underserved regions.

From a health economics perspective, artificial intelligence may reduce costs by optimizing diagnostic pathways, minimizing unnecessary imaging and outpatient visits, and preventing advanced-stage complications. Early risk identification and timely intervention may reduce the need for intensive therapies and long-term disability-related expenditures.

From a broader policy and healthcare systems perspective, the findings of this bibliometric analysis may help inform strategic planning for the integration of AI into lymphedema care pathways. Policymakers and healthcare administrators should consider the need for regulatory frameworks, clinician training programs, digital infrastructure investment, and reimbursement strategies to facilitate the safe and equitable implementation of AI-supported technologies in clinical practice.

Despite its considerable clinical potential, the implementation of artificial intelligence in lymphedema care also raises important ethical and practical challenges. AI models may be affected by algorithmic bias and dataset imbalance, potentially limiting fairness and generalizability across diverse patient populations. In addition, the use of large clinical and imaging datasets raises concerns regarding data privacy, cybersecurity, and patient confidentiality. Model explainability and transparency remain essential to ensure clinician trust and facilitate responsible integration into clinical workflows. Regulatory oversight and ethical governance frameworks will therefore be critical to support the safe, equitable, and accountable implementation of AI technologies in healthcare settings.

Overall, artificial intelligence supports a transition toward personalized, proactive, and data-driven lymphedema care. Although current evidence remains limited by small sample sizes and heterogeneous methodologies, existing findings indicate substantial clinical potential. Future research should prioritize prospective validation, multicenter collaboration, and standardized reporting to facilitate safe clinical implementation.

Limitations

Several limitations should be acknowledged. First, the relatively limited number of included studies may restrict the generalizability of the findings. However, this likely reflects the emerging and highly specialized nature of artificial intelligence applications in lymphedema research rather than a methodological limitation of the present study. Citation-based analyses inherently favor older publications due to cumulative citation accrual over time, which may lead to the relative underrepresentation of recently published studies despite their potential scientific relevance. Additionally, the predominance of publications between 2022 and 2025 may reflect an early expansion phase of the field rather than stable long-term research trends. As a bibliometric and altmetric study, methodological quality, risk-of-bias assessment, and direct evaluation of clinical validity were beyond the intended scope of the present analysis. Furthermore, the restriction to English-language publications indexed in the Web of Science Core Collection and the exclusion of other databases may have limited the comprehensiveness of the dataset. A detailed analysis of altmetric subcomponents was not performed, which may have limited more granular interpretation of digital dissemination patterns. These factors may limit the comprehensive representation of interdisciplinary research dynamics.

Conclusion

This study offers a concise overview of current research trends on the use of artificial intelligence in lymphedema. Findings emphasize the prominence of diagnosis, risk prediction, and imaging analysis. The notable rise in publications in recent years highlights increasing academic interest and interdisciplinary collaboration. To support a more comprehensive understanding, future research should integrate broader databases, multilingual sources, and detailed altmetric evaluations.

Supplemental material

Supplemental Material - Emerging trends in artificial intelligence research in lymphedema: An evaluation in light of bibliometric and altmetric data

Supplemental Material for Emerging trends in artificial intelligence research in lymphedema: An evaluation in light of bibliometric and altmetric data by Nurmuhammet Taş, Gürhan Çelik, Yakup Erden, Mustafa Hüseyin Temel, Fatih Bağcıer and Evrim Coşkun in Phlebology.

Footnotes

ORCID iDs

Nurmuhammet Taş

Mustafa Hüseyin Temel

Author contributions

Nurmuhammet Taş: Methodology, Software, Validation, Formal analysis, Data curation, Writing – original draft. Gürhan Çelik: Conceptualization, Writing – review & editing, Supervision, Project administration, Funding acquisition. Yakup Erden: Writing – review & editing. Mustafa Hüseyin Temel: Investigation. Fatih Bağcıer: Investigation, Resources, Software, Writing – review & editing, Conceptualization. Evrim Coskun: Investigation, Resources, Software. All authors reviewed and approved the final manuscript. All authors have read and approved submission of the manuscript and the manuscript has not been published and is not being considered for publication elsewhere in whole or part in any language.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Guarantor

Corresponding author: Dr. Nurmuhammet Taş is the guarantor of this article and accepts full responsibility for the work, had access to the data, and controlled the decision to publish.

Supplemental material

Supplemental material for this article is available online.

References

Zhang

Wang

, et al. Visual analysis of global research output of lymphedema based on bibliometrics. Front Oncol 2022; 12: 926237. https://doi.org/10.3389/fonc.2022.926237

DiSipio

Rye

Newman

, et al. Incidence of unilateral arm lymphoedema after breast cancer: a systematic review and meta-analysis. Lancet Oncol 2013; 14(5): 500–515. https://doi.org/10.1016/S1470-2045(13)70076-7

Trinh

Chien

Long

, et al. Development of predictive models for lymphedema by using blood tests and therapy data. Sci Rep 2023; 13(1): 19720. https://doi.org/10.1038/s41598-023-46567-1

Szuba

Rockson

. Lymphedema: classification, diagnosis and therapy. Vasc Med 1998; 3(2): 145–156. https://doi.org/10.1177/1358836X9800300209

Shang

Zhou

Pan

, et al. Bibliometric and visual analysis of breast cancer-related lymphedema: knowledge structure, research status, and future trends. Medicine 2025; 104(7): e41510. https://doi.org/10.1097/md.0000000000041510

Sakorafas

Peros

Cataliotti

, et al. Lymphedema following axillary lymph node dissection for breast cancer. Surg Oncol 2006; 15(3): 153–165. https://doi.org/10.1016/j.suronc.2006.11.003

Rosedale

. Breast cancer survivors’ experiences of lymphedema-related symptoms. J Pain Symptom Manag 2009; 38(6): 849–859. https://doi.org/10.1016/j.jpainsymman.2009.04.030

Bucak

ÖF

. Unveiling global interest in lymphedema: a decade of search data through google trends. Phlebology 2024; 39(10): 2683555241310749.

Lurie

Malgor

Carman

, et al. The American venous forum, American vein and lymphatic society and the society for vascular medicine expert opinion consensus on lymphedema diagnosis and treatment. Phlebology 2022; 37(4): 252–266. https://doi.org/10.1177/02683555211053532

10.

Donahue

PMC

MacKenzie

Filipovic

, et al. Advances in the prevention and treatment of breast cancer-related lymphedema. Breast Cancer Res Treat 2023; 200(1): 1–14. https://doi.org/10.1007/s10549-023-06947-7

11.

Cornish

Chapman

Hirst

, et al. Early diagnosis of lymphedema using multiple frequency bioimpedance. Lymphology 2001; 34(1): 2–5.

12.

Ruehm

Schroeder

Debatin

. Interstitial MR lymphography with gadoterate meglumine: initial experience in humans. Radiology 2001; 220(3): 816–821. https://doi.org/10.1148/radiol.2203010090

13.

Choi

, et al. Automated CT segmentation for lower extremity tissues in lymphedema evaluation using deep learning. Eur Radiol 2025. Epub ahead of print.

14.

Armer

Radina

Porock

, et al. Predicting breast cancer-related lymphedema using self-reported symptoms. Nurs Res 2003; 52(6): 370–379. https://doi.org/10.1097/00006199-200311000-00004

15.

Davenport

Kalakota

. The potential for artificial intelligence in healthcare. Future Healthc J 2019; 6(2): 94–98. https://doi.org/10.7861/futurehosp.6-2-94

16.

Chen

Xie

Tao

, et al. Artificial intelligence and multimodal data fusion for smart healthcare: topic modeling and bibliometrics. Artif Intell Rev 2024; 57: 91. https://doi.org/10.1007/s10462-024-10712-7

17.

Topol

. High-performance medicine: the convergence of human and artificial intelligence. Nat Med 2019; 25(1): 44–56. https://doi.org/10.1038/s41591-018-0300-7

18.

Esteva

Kuprel

Novoa

, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature 2017; 542(7639): 115–118. https://doi.org/10.1038/nature21056

19.

Lin

, et al. A bibliometric analysis of the advance of artificial intelligence in medicine. Front Med 2025; 12: 1504428. https://doi.org/10.3389/fmed.2025.1504428

20.

Wei

Jin

, et al. Developing and validating a prediction model for lymphedema detection in breast cancer survivors. Eur J Oncol Nurs 2021; 54: 102023. https://doi.org/10.1016/j.ejon.2021.102023

21.

Bağcıer

İnal

Çağlar

. Top 100 cited articles on fibromyalgia syndrome: a bibliometric and altmetric analyses study. Agri 2021; 33(2): 53–63.

22.

Bian

Shen

Yang

, et al. Financial toxicity experienced by patients with breast cancer-related lymphedema: a systematic review. Support Care Cancer 2023; 31(6): 354. https://doi.org/10.1007/s00520-023-07800-9

23.

Dean

Moss

Ransome

, et al. “It still affects our economic situation”: long-term economic burden of breast cancer and lymphedema. Support Care Cancer 2019; 27: 1697–1708. https://doi.org/10.1007/s00520-018-4418-4

24.

Brayton

Hirsch

O’Brien

, et al. Lymphedema prevalence and treatment benefits in cancer: impact of a therapeutic intervention on health outcomes and costs. PLoS One 2014; 9(12): e114597. https://doi.org/10.1371/journal.pone.0114597

25.

Lewsirirat

Titijaroonroj

Apichonbancha

, et al. Deep learning-based classification of lymphedema and other lower limb edema diseases using clinical images. Sci Rep 2025; 15(1): 12453. https://doi.org/10.1038/s41598-025-97564-5

26.

Eldaly

Avila

Guzman

RAT

, et al. Artificial intelligence and lymphedema: state of the art. J Clin Transl Res 2022; 8(3): 234–242.

27.

Donthu

Kumar

Mukherjee

, et al. How to conduct a bibliometric analysis: an overview and guidelines. J Bus Res 2021; 133: 285–296. https://doi.org/10.1016/j.jbusres.2021.04.070

28.

Ayoub

Amin

Wani

. Exploring the impact of altmetrics in relation to citation count and SCImago journal rank (SJR). J Scientometr Res 2023; 12(3): 603–608. https://doi.org/10.5530/jscires.12.3.058

29.

Thelwall

Haustein

Larivière

, et al. Do altmetrics work? Twitter and ten other social web services. PLoS One 2013; 8(5): e64841. https://doi.org/10.1371/journal.pone.0064841

30.

Agarwal

Durairajanayagam

Tatagari

, et al. Bibliometrics: tracking research impact by selecting the appropriate metrics. Asian J Androl 2016; 18(2): 296–309. https://doi.org/10.4103/1008-682X.171582

31.

Kistenev

Nikolaev

Kurochkina

, et al. Application of multiphoton imaging and machine learning to lymphedema tissue analysis. Biomed Opt Express 2019; 10(7): 3353–3368. https://doi.org/10.1364/BOE.10.003353

32.

Goudarzi

Whyte

Boily

, et al. Segmentation of arm ultrasound images in breast cancer-related lymphedema: a database and deep learning algorithm. IEEE Trans Biomed Eng 2023; 70(9): 2552–2563. https://doi.org/10.1109/tbme.2023.3253646

33.

Fallahzadeh

Pedram

Saeedi

, et al. Smart-cuff: a wearable bio-sensing platform with activity-sensitive information quality assessment for monitoring ankle edema. In: 2015 IEEE international conference on pervasive computing and communication workshops (percom workshops), St. Louis, MO, 23–27 March 2015.

34.

Rezende

LFD

Piloni

JPM

Kempa

, et al. Ultrasonography as an instrument to evaluate lymphedema secondary to breast cancer: systematic review. J Vasc Bras 2023; 18(22): e20220144. https://doi.org/10.1590/1677-5449.202201442

35.

Wang

Nie

Zhu

, et al. Artificial intelligence assisted ultrasound for the non-invasive prediction of axillary lymph node metastasis in breast cancer. BMC Cancer 2024; 24: 910. https://doi.org/10.1186/s12885-024-12619-6

36.

Teo

JYC

Wang

. Interdisciplinary collaboration for chronic illness prevention and management through digital health interventions. Interdiscip Nurs Res 2024; 3(4): 199–200. https://doi.org/10.1097/nr9.0000000000000073

37.

Jayasree

Vijayakumar

Sugumaran

, et al. A comprehensive review of computational diagnostic techniques for lymphedema. Prog Biomed Eng 2025; 7(2): 022002.

38.

Wang

, et al. Machine learning for detection of lymphedema among breast cancer survivors. mHealth 2018; 4: 17. https://doi.org/10.21037/mhealth.2018.04.02

39.

Yang

, et al. Comparison of machine learning models to predict the risk of breast cancer-related lymphedema among breast cancer survivors: a cross-sectional study in China. Front Oncol 2024; 14: 1334082. https://doi.org/10.3389/fonc.2024.1334082

40.

Huang

, et al. Bibliometric analysis of breast cancer-related lymphedema research trends over the last 2 decades. Front Oncol 2024; 14: 1360899. https://doi.org/10.3389/fonc.2024.1360899

41.

Zahedi

Haustein

. On the relationships between bibliographic characteristics of scientific documents and citation and Mendeley readership counts: a large-scale analysis of web of science publications. J Informetr 2018; 12(1): 191–202. https://doi.org/10.1016/j.joi.2017.12.005

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.21 MB

0.00 MB