The study, led by Wenyu Song, Siyu Gong, and Zhigang Yu, addresses a gap that has grown as rapidly as the literature itself. While publications on agricultural resilience have surged in recent years, driven by intensifying climate volatility and mounting concerns over global food systems, few efforts have systematically assessed the field’s structure, key contributors, and thematic trajectories. Bibliometric analysis, which applies quantitative methods to publication data, provides a way to see the forest rather than the trees. By combining CiteSpace, which excels at detecting emerging trends and conceptual turning points through keyword co-occurrence and burst detection, with SciMAT, which constructs longitudinal thematic maps that track how research themes evolve across successive time periods, the team was able to capture both the fine-grained dynamics and the long-term structural shifts in the field.

The analysis reveals that agricultural resilience research has progressed through three distinct evolutionary stages. The first, spanning roughly the early years of the study window, was characterized by initial theoretical exploration. During this period, researchers focused on defining what resilience means in an agricultural context, borrowing heavily from ecological and systems theory. Concepts such as adaptive capacity, buffer capacity, and the ability of farming systems to maintain essential functions under disturbance dominated the discourse. This foundational phase was essential for establishing the conceptual vocabulary that would later underpin empirical work, but it remained largely abstract and disconnected from on-the-ground agricultural challenges.

The second stage marked a decisive turn toward practical application. As climate extremes became more frequent and severe, researchers began translating theoretical resilience concepts into measurable indicators and actionable strategies. Studies increasingly examined how specific farming systems, from smallholder operations in developing regions to large-scale commercial enterprises, respond to droughts, floods, pest outbreaks, and market shocks. This period saw the rise of case-based research, regional vulnerability assessments, and the first attempts to quantify resilience using observable variables such as crop yield stability, income diversification, and soil health metrics. The shift from theory to practice reflected a growing recognition that resilience is not merely an academic construct but a tangible property that can be strengthened through targeted interventions.

The third and most recent stage is characterized by integrative studies that seek to synthesize insights across disciplines, scales, and methodologies. Current research, according to the keyword analysis, focuses on adaptability, ecosystem services, and integrated analytical frameworks. This convergence reflects an understanding that agricultural resilience cannot be understood in isolation from the broader social-ecological systems in which farms are embedded. Ecosystem services, the benefits that natural environments provide to agricultural production, such as pollination, water regulation, and soil fertility, have emerged as a central theme, linking resilience thinking to conservation biology, landscape ecology, and environmental economics. At the same time, the integration of technological innovations, including remote sensing, big data analytics, and machine learning, is opening new avenues for monitoring and predicting system behavior under stress.

Despite this impressive trajectory, the bibliometric analysis identifies significant challenges that continue to constrain the field. Chief among them is the absence of standardized measurement indicators. Unlike fields such as epidemiology or economics, where established metrics allow for direct comparison across studies and contexts, agricultural resilience research lacks a universally accepted set of indicators. Different studies employ different proxies, scales, and analytical units, making it difficult to synthesize findings or build cumulative knowledge. This fragmentation hampers the development of robust predictive models and limits the ability of policymakers to draw evidence-based conclusions about which interventions are most effective.

The second major challenge is pronounced regional imbalance in research attention. The analysis of national research output and collaboration networks reveals that agricultural resilience scholarship is heavily concentrated in a small number of countries, with developed nations and major emerging economies dominating publication counts and citation networks. Regions that are arguably most vulnerable to climate shocks, including parts of sub-Saharan Africa, South Asia, and small island developing states, remain underrepresented in the literature. This geographic skew not only reflects existing inequalities in research funding and infrastructure but also risks producing knowledge that is poorly suited to the specific climatic, cultural, and socioeconomic conditions of the most exposed populations. The authors argue that addressing this imbalance is essential for developing adaptation strategies that are both scientifically rigorous and locally relevant.

Looking forward, the study proposes several priorities for advancing the field. First, fostering interdisciplinary collaboration is critical. Agricultural resilience sits at the intersection of agronomy, ecology, economics, sociology, and climate science, yet much of the existing literature remains siloed within disciplinary boundaries. Second, integrating technological innovation with policy analysis could bridge the gap between what is technically possible and what is institutionally feasible. Precision agriculture, early warning systems, and climate-smart farming practices hold enormous promise, but their adoption depends on supportive policies, market incentives, and farmer capacity. Third, the development of multi-scale assessment frameworks is needed to capture resilience dynamics that operate simultaneously at the plot, farm, landscape, regional, and global levels. Such frameworks would enable researchers to identify trade-offs and synergies across scales, informing more coherent and effective policy design.

The implications of this research extend well beyond academia. As climate uncertainty intensifies and global food systems face unprecedented pressure from population growth, resource degradation, and geopolitical instability, understanding how agricultural systems can withstand and adapt to shocks becomes a matter of global security. The bibliometric map produced by Song, Gong, and Yu provides researchers, funders, and policymakers with a navigational tool for identifying where the field has been, where it stands, and where it needs to go. By revealing the intellectual architecture of agricultural resilience research, this analysis lays the groundwork for a more coordinated, inclusive, and impactful research agenda capable of guiding region-specific adaptation strategies in an increasingly unpredictable world.

Subject of Research: Bibliometric mapping of the evolution and knowledge structure of agricultural resilience research

Article Title: Multidimensional evolutionary paths in agricultural resilience research: a bibliometric analysis based on citespace and SciMAT

Article References: Song, W., Gong, S., & Yu, Z. (2026). Multidimensional evolutionary paths in agricultural resilience research: a bibliometric analysis based on citespace and SciMAT. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-026-09181-w

Image Credits: AI Generated

DOI: 10.1057/s41599-026-09181-w

Keywords: agricultural resilience, bibliometrics, CiteSpace, SciMAT, food security, climate adaptation, ecosystem services, sustainability science, research trends, interdisciplinary collaboration, knowledge mapping, Northeast Agricultural University

The study, conducted by Mengrui Zhu of Xi’an Jiaotong University and Mengjiao Li of Gao Xin No. 1 Lugang Secondary School in Xi’an, China, is not a conventional review. Instead of summarizing individual papers, the researchers applied bibliometrics, the quantitative mapping of scientific literature, to identify the structural anatomy of an entire research domain. Drawing on the Web of Science Core Collection, they retrieved 245 records using a carefully constructed topic search that combined terms for video and online formats, instructor and teacher roles, second language and L2 contexts, and learning. After duplicate checks and title-and-abstract screening by two researchers with backgrounds in English language teaching and second language acquisition, 211 journal articles were retained for analysis. The corpus spanned 72 journals, 454 authors, 395 institutions, and 91 regions, providing a broad international snapshot of scholarship on the teacher behind the screen.

To visualize the knowledge landscape, the team used CiteSpace, version 6.1.R6, a software platform developed by Chaomei Chen’s team at Drexel University that is widely used for detecting research hotspots, intellectual structures, and emerging trends in scientific literature. Keywords served as the primary analytical nodes, selected with the default g-index criterion set to k = 25. The keyword co-occurrence and clustering network was generated using one-year time slices and Pathfinder pruning, while a separate time-zone map was produced with three-year time slices and minimum spanning tree pruning, configurations chosen to sharpen the clarity of thematic and temporal structures. The resulting co-occurrence network contained 340 nodes and 1,133 links, with a modularity value of Q = 0.6279, well above the reliability threshold of 0.3, and a weighted mean silhouette value of S = 0.8437, exceeding the recommended level of 0.5. In plain terms, the statistical diagnostics confirmed that the thematic clusters the algorithm detected were both significant and trustworthy, not artifacts of the software’s parameters.

The publication trajectory itself tells a dramatic story. In the years immediately following 2007, research output remained persistently low, never exceeding five articles per year, indicating that the video instructor was still an exploratory topic with no sustained publication momentum. After 2014, output began to climb, though unevenly: twelve articles appeared in 2017, the count collapsed to just three in 2018, and then rebounded to fourteen in 2019. Then came the inflection point. Beginning in 2020, coinciding with the COVID-19 pandemic’s forced migration of language classrooms online, publications surged to eighteen articles, rose to twenty-eight in 2021, and peaked at thirty-three in 2022. Although output moderated to thirty articles in 2023 and twenty-six in 2024, it remained far above pre-pandemic levels, suggesting that the video instructor has evolved from an emergency measure into a permanent fixture of second language education research.

By integrating the network structure with clustering results, the analysis identified four major research themes that now define the field. The first concerns the evolution of technologies and methods, encompassing clusters such as blended learning, computer-mediated communication, computer-assisted language learning, and artificial intelligence, effectively charting the technological maturation of language teaching from auxiliary tool to central medium. The second theme centers on learner experience and affective factors, dominated by clusters on second language boredom and autonomy, highlighting how motivation, emotion, and self-regulation shape learning in video-mediated environments. The third theme involves the expansion of linguistic and cultural literacies, including multiliteracy, first language use, and foreign language learning, reflecting a shift from narrow skill training toward meaning construction and intercultural competence in digital and multimodal settings. The fourth theme focuses on language interaction and discourse analysis, with conversation analysis at its core, underscoring the crucial role of interactional processes in language acquisition and meaning construction.

The affective theme may be the most striking, because it reveals how the science of screen-based teaching has converged with the psychology of emotion. Early research organized around keywords such as student and classroom examined how interaction patterns and instructional behaviors influenced participation, viewing classroom interaction as a process co-constructed through questioning, feedback, and responsive exchange rather than one-directional transmission. After 2020, keywords such as boredom, anxiety, and engagement appeared with mounting frequency. Empirical studies consistently show that learners in video-based classrooms often experience boredom, fatigue, and feelings of disconnection, which undermine motivation and reduce participation. Crucially, the sources of these emotions are contextual rather than personal: monotonous instructional formats, insufficient interaction, and problems with technology and classroom organization are the major drivers of boredom in online environments. Learners, the evidence suggests, expect video instructors to enhance interaction opportunities, optimize activity design, and adjust pacing, indicating that emotional experience in online classrooms is inseparable from instructional organization and interaction quality.

Burst detection, a technique proposed by Kleinberg that identifies terms experiencing rapid increases in frequency, added a temporal dimension to the mapping. Early burst keywords concentrated on online, knowledge, teacher education, conversation analysis, and computer-assisted language learning, reflecting foundational work on technology integration. Conversation analysis exhibited a long burst beginning in 2012, testifying to its sustained methodological importance in dissecting interactional structures, teacher questioning and feedback practices, and learner participation patterns in video-mediated instruction. Around 2020, the burst landscape shifted decisively toward the learner: perception and performance emerged between 2020 and 2021, followed by sustained bursts in anxiety, motivation, positive psychology, and enjoyment. After 2022, keywords such as strategy, belief, learner, classroom, and feedback signaled a more holistic view, conceptualizing video-based instruction within a broader learning ecology in which learners’ beliefs and strategies, teachers’ interactional practices, and feedback mechanisms jointly shape participation and emotional experience.

The time-zone visualization, generated with 225 nodes and 300 links, delineated the field’s stage-by-stage evolution with remarkable precision. From 2007 to 2009, research focused on traditional classroom interaction and language comprehension, with keywords such as teacher-student interaction, corrective feedback, and comprehension prominent, and video used mainly for classroom recording in face-to-face contexts. Between 2010 and 2012, attention extended toward online learning, distance education, and peer teaching, while conversation analysis and teacher education entered the landscape. From 2013 to 2015, classroom discourse, gesture in language learning, and speech dominated, reflecting growing interest in visual input and multimodal behavior. Between 2016 and 2018, the focus moved to the learner level with achievement, adult learners, and form. From 2019 to 2021, motivation, anxiety, performance, and willingness to communicate were highly concentrated, and from 2022 to 2024, cognitive engagement, listening comprehension, and teacher professional development signaled attention to cognitive processing and teaching optimization. In the partial 2025 data, cognitive theory and communication breakdown appeared as preliminary hints of emerging interests.

The knowledge structure that emerges from the analysis can be summarized across three interrelated levels: the technological conditions constituted by video platforms, multimodal presentation, and artificial intelligence tools; the instructional process centered on teacher-student interaction and affective regulation; and the level of learning experience and outcomes, focusing on learners’ cognitive and affective processing and performance. Evidence from individual empirical studies suggests that teacher demonstrations, visual cues, and affective expression may be associated with lower cognitive load, more focused attention, and improved learning attitudes under specific instructional conditions. In one particularly compelling neurocognitive study, the visible presence of an instructor in instructional videos was associated with greater neural synchrony and visual attention during multimedia learning, suggesting that the human face on screen literally shapes how brains process educational content. For practice, the structure implies that effective video-based language teaching must integrate technological choices, interaction design, affective support, and cognitive load management rather than relying on isolated tools or strategies.

The authors acknowledge limitations: the data came exclusively from the Web of Science Core Collection, only English-language journal publications were included, and the search strategy was restricted to studies explicitly using the terms second language or L2, potentially underrepresenting research framed in terms of L3, heritage language, foreign language, or multilingual learning. The 2025 data covered only the period through July 9, and clustering results were partly influenced by CiteSpace parameter settings and keyword standardization procedures. Yet the frontier questions the study identifies are unmistakably forward-looking. Future research, the authors argue, should compare human instructors, prerecorded video instructors, digital humans, and generative AI instructors to determine whether they differ in attracting learner attention, conveying emotional information, and supporting classroom interaction, and whether these differences relate to cognitive load, engagement, and outcomes across learners. Such questions could be addressed through behavioral experiments, eye-tracking, event-related potentials, longitudinal research, and cross-cultural comparisons. As artificial intelligence tools continue to be incorporated into video instruction, maintaining interpersonal interaction and affective support while leveraging technological affordances has become the defining challenge, marking the field’s shift from a tool-oriented perspective to a mechanism-oriented science of how languages are learned through screens.

Subject of Research: Bibliometric mapping of research on video and online instructors in second language learning from 2007 to 2025

Article Title: Instructors in video-based and online second language learning: a bibliometric analysis (2007–2025)

Article References: Zhu, M., & Li, M. (2026). Instructors in video-based and online second language learning: a bibliometric analysis (2007–2025). Discover Education, 5(1), Article 886. https://doi.org/10.1007/s44217-026-02004-9

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02004-9

Keywords: video instructors, second language learning, bibliometric analysis, CiteSpace, online language teaching, learner engagement, computer-assisted language learning, artificial intelligence in education, classroom interaction, affective factors, Instructors, video-based

The scale of the undertaking is itself remarkable. The researchers searched five major databases simultaneously: Web of Science and PubMed for the international, English-language literature, and CNKI, WanFang, and CQVIP for the vast Chinese-language corpus that most Western observers never see. This dual-language approach matters enormously, because the overwhelming majority of research on Chinese patent medicines has been conducted in mainland China and published in Chinese journals. By preprocessing the data with Co-Occurrence 20.5 and then analyzing it with CiteSpace and R software, the team could visualize co-citation networks, keyword bursts, and institutional collaborations — essentially building a cartographic map of an entire field, with its mountains of activity, its valleys of neglect, and its emerging frontier ridges.

What the map shows is a field that has grown steadily, with annual publication output concentrated overwhelmingly in mainland China. That geographic concentration is not merely a curiosity of publishing patterns; it raises fundamental questions about the universality of the evidence base. If nearly all studies of a therapy come from one country’s healthcare system, with its particular patient populations, treatment protocols, and regulatory environment, how confidently can the findings be generalized to the diverse, multiethnic patient populations who increasingly turn to these products worldwide? The authors explicitly flag this problem, calling for more cross-regional and multiethnic studies to validate the universal applicability and safety of Chinese patent medicines before any broader use can be considered.

Perhaps the most striking finding concerns the gap between what clinical guidelines recommend and what the products’ own package inserts say. Among the top ten Chinese patent medicines with the highest volume of research, most had received recommendations from at least one clinical practice guideline — documents from bodies such as the China Anti-Cancer Association, the National Health Commission of China, and various expert consensus panels on integrated traditional Chinese and Western medicine. Yet those recommendations sometimes diverged from the package inserts, which often lacked explicit tumor indications altogether. In other words, oncologists and traditional Chinese medicine practitioners may be prescribing these medicines for cancer-related purposes that the official labeling never anticipated, let alone validated through rigorous clinical trials.

This discrepancy between actual clinical application and usage standards is the study’s central alarm bell. It is not a new problem in complementary medicine, but it is unusually acute here because Chinese patent medicines occupy a peculiar regulatory and cultural position: they are standardized pharmaceutical products, manufactured under quality controls, yet their therapeutic logic derives from a medical framework — pattern differentiation and formula composition — that does not map neatly onto modern oncology’s molecular categories. The authors argue that closing this gap requires clarifying the drugs’ mechanisms and clinical value through high-quality evidence, a process they describe as essential to the scientific and standardization development of the field.

The bibliometric analysis also exposed a fascinating linguistic and philosophical divide within the literature itself. Studies published in English-language databases tended to focus on Chinese patent medicines prepared or purified from only one or two herbal ingredients, with clearly identified primary active compounds and botanical origins. Think of β-elemene, the sesquiterpene from Curcuma species shown in laboratory studies to reverse chemoresistance in breast cancer cells by disrupting exosome-mediated resistance transmission, or astragaloside IV, the saponin from Astragalus membranaceus that enhanced paclitaxel chemosensitivity through caveolin-1-targeted oxidant damage. These are molecules a Western pharmacologist can immediately grasp: a defined compound, a defined target, a defined mechanism.

Chinese-language research, by contrast, embraces the full complexity of multi-herb formulations — the classic compound prescriptions like Xihuang Pill, a centuries-old formula whose antitumor effects in mouse models of breast cancer have been linked to reductions in regulatory T cells within the tumor microenvironment via the PI3K/AKT–AP-1 signaling pathway. This divide is more than stylistic. Single-compound agents are far easier to subject to randomized controlled trials, pharmacokinetic analysis, and regulatory scrutiny, while multi-herb formulas resist the reductionist toolkit but may capture synergistic interactions that single agents miss. The field’s future trajectory may well depend on whether these two research cultures can find a common methodological language.

Where is the intellectual energy of the field heading? The analysis identified two dominant research hotspots: the tumor microenvironment, and the strategy of reducing toxicity while enhancing efficacy in conventional antitumor therapies — the concept Chinese oncologists call ‘jian du zeng xiao.’ The first hotspot reflects mainstream cancer biology’s own migration. Chronic inflammation, immunosuppression, tumor-associated macrophages, and regulatory T cells are now recognized as central players in breast cancer progression, and traditional Chinese medicine researchers have positioned their agents as modulators of this microenvironmental ecology. The second hotspot addresses a persistent clinical pain point: chemotherapy-induced leukopenia, nausea, cancer-related fatigue, cancer pain, and cardiotoxicity drive dose reductions and treatment discontinuation, and several expert consensus documents in China now formally recommend Chinese patent medicines for managing these toxicities.

One rising star deserves particular attention: herbal polysaccharides. According to the study’s trend analysis, these complex carbohydrates — long dismissed as pharmacologically inert ‘fiber’ — are gaining both attention and academic influence within the field, appearing to be the current wave of scholarly interest. Polysaccharides from mushrooms and roots such as Astragalus have been investigated for immunomodulatory properties, and their growing bibliometric footprint suggests researchers see them as a promising interface between traditional formulations and modern immuno-oncology. Whether that promise survives contact with rigorous clinical evaluation remains to be seen, but the trend line is unmistakable.

For patients and clinicians, the study’s message is nuanced. Chinese patent medicines are widely used in cancer care, embedded in Chinese clinical guidelines, and supported by a substantial and growing research literature — 1,847 publications is not nothing. Yet the evidence architecture is uneven: recommendations sometimes outpace labeling, package inserts often lack tumor indications, the research base is geographically concentrated, and the highest-quality clinical validation remains the exception rather than the rule. The authors’ prescription is clear and measured: improve scientific rigor and standardization, clarify mechanisms and clinical value with high-quality evidence, and expand studies across regions and ethnic populations. Until then, the landscape they have mapped remains a terrain of genuine promise — and genuine uncertainty — where the gap between what is practiced and what is proven is the single most important feature on the map.

Subject of Research: Bibliometric mapping of Chinese patent medicines research in breast cancer care from 1991 to 2025

Article Title: Mapping the landscape of Chinese patent medicines in breast cancer: a multi-database bibliometric analysis (1991-2025)

Article References: Yuan, Y., Ji, Y., Wang, Z.-J., Xia, X.-P., Yang, C.-Y., Huang, R.-Y., Zhang, Y.-J., Zuo, J.-L., Liu, X.-Q., & Liu, J.-P. (2026). Mapping the landscape of Chinese patent medicines in breast cancer: a multi-database bibliometric analysis (1991-2025). Supportive Care in Cancer, 34(10), Article 967. https://doi.org/10.1007/s00520-026-11209-5

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11209-5

Keywords: breast cancer, Chinese patent medicines, traditional Chinese medicine, bibliometric analysis, complementary therapy, tumor microenvironment, chemotherapy side effects, herbal polysaccharides, clinical guidelines, evidence-based medicine, CiteSpace, supportive cancer care

The team used CiteSpace, a widely adopted bibliometric software tool developed by Chaomei Chen, to visualize the intellectual landscape of the field. Rather than reading every paper in depth, bibliometric analysis treats the literature itself as data: publication counts, co-authorship links between countries and institutions, keyword co-occurrence patterns, thematic clusters, and sudden surges in the popularity of specific terms, known as citation bursts. This approach allows researchers to detect the invisible architecture of a scientific discipline, revealing which ideas dominate, which are fading, and which are emerging. The authors are careful to note an important caveat: these patterns reflect how the research community has organized itself and where it has chosen to focus, not direct evidence of the underlying climatological mechanisms themselves.

One of the clearest findings is a dramatic acceleration in publication output, particularly after 2015. This timing is no accident. The past decade has seen a string of deadly heatwaves across Europe, South Asia, and the Middle East, alongside rapid urbanization in China, Southeast Asia, and Africa. As more than half of the global population now lives in cities, a figure highlighted by United Nations urbanization projections, the stakes of understanding urban heat have risen sharply. Previous studies have attributed excess mortality during heatwaves to the urban heat island effect, and researchers have increasingly recognized that the geometry of streets, the density of buildings, and the arrangement of green space are not merely aesthetic choices but determinants of thermal exposure for millions of people.

The collaboration analysis reveals a field that is geographically concentrated but only partially integrated. China and the United States emerge as the dominant publication and collaboration nodes, with several European and Asian research institutions forming additional major hubs. Yet the networks remain fragmented: substantial portions of the literature are produced within national or institutional silos, with limited cross-border integration. This matters because urban climate problems are inherently comparative. A street canyon in Hong Kong behaves differently from one in Phoenix or Karlsruhe, and the field’s ability to generalize depends on researchers in different climates and planning cultures testing each other’s methods and findings. The authors identify this fragmentation as a key weakness that future work must address.

On the thematic side, the keyword analysis identified a stable core of persistent research concerns. Land surface temperature and air temperature remain the fundamental measured variables, typically retrieved from thermal remote sensing satellites or dense networks of ground sensors. Sky view factor, a geometric measure of how much of the sky is visible from a given point on the ground, and street-canyon geometry recur constantly because they govern both the trapping of longwave radiation at night and the shading and airflow that moderate daytime temperatures. Thermal comfort, the human-centered dimension of the problem, also features prominently, connecting the physics of urban climates to the experience of pedestrians.

The temporal and burst analyses tell a story of cumulative broadening. In the earlier years of the study period, research emphasized foundational measurement and the morphological characterization of urban surfaces: cataloging building densities, impervious surface fractions, and vegetation indices, and linking them to satellite-derived temperature patterns. Over time, while these foundations remained visible in the literature, computational modeling gained increasing prominence. Researchers moved from describing correlations to simulating the physical processes involved, using urban canopy models, computational fluid dynamics, and energy balance approaches to predict how specific configurations of buildings and vegetation alter airflow, turbulence, and heat removal from street canyons. Laboratory techniques such as water tunnel measurements with particle image velocimetry have even been brought to bear on the question of how buoyant flows flush heat out of urban canyons.

Two conceptual frameworks stand out in the field’s recent evolution. The first is the Local Climate Zone classification system, introduced by Ian Stewart and Tim Oke in 2012, which provides a standardized taxonomy of urban surface types, from compact high-rise to scattered low-rise and dense trees. The review found that Local Climate Zones have become an increasingly prominent organizing tool, allowing researchers in Nanjing, Hong Kong, Phoenix, and Las Vegas to compare heat island behavior across structurally similar but climatically different neighborhoods. The second is the rise of outdoor thermal comfort as a planning-oriented endpoint. Studies of pocket parks in high-rise high-density environments, street configurations in Mediterranean climates, and human-centered design strategies for dense urban spaces signal a shift from diagnosis toward intervention.

This shift toward application is perhaps the most consequential trend the review documents. The literature increasingly connects urban morphology to questions of policy and practice: how planning indicators shape heat island intensity, how compactness and green infrastructure strategies trade off against ventilation and energy consumption, and how building energy demand is coupled to the thermal environment that urban form creates. The authors note that planning-oriented applications have gained increasing prominence in the most recent phase of the field, reflecting demand from city governments for actionable guidance as heat action plans become standard instruments of climate adaptation. Lessons from policy implementation in multiple cities are now entering the peer-reviewed record, closing the loop between research and practice.

At the same time, the review is candid about the limits of what bibliometrics can show. The patterns it reveals describe the organization and priorities of the research literature, not the causal structure of urban climates themselves. A surge in keywords related to thermal comfort indicates where scholarly attention has flowed, not necessarily where the most important physical mechanisms lie. The authors argue that this distinction matters for interpreting the field’s trajectory and for identifying genuine gaps. Among the areas they highlight as requiring more integrated investigation are comparative studies spanning multiple cities and climates, methodological work to harmonize the diverse measurement and modeling approaches currently in use, and application-oriented research that rigorously tests whether design interventions deliver the cooling benefits they promise.

For a field born from the observation that cities make their own weather, the past two decades have transformed urban heat research from a descriptive exercise into a quantitative, interdisciplinary, and increasingly policy-relevant science. This new knowledge map provides the field with something it has lacked: a systematic account of its own growth, its geographic asymmetries, and its thematic evolution. As heatwaves intensify and urban populations expand, the geometry of the built environment will remain one of the few levers cities can actually pull to protect their residents from extreme heat. Knowing precisely where the science stands, and where the silos and gaps persist, is the first step toward pulling that lever effectively.

Subject of Research: Bibliometric analysis of urban morphology and urban heat island research from 2005 to 2024

Article Title: Urban morphology and the urban heat island effect: a bibliometric review and knowledge mapping of research evolution and thematic trends (2005–2024)

Article References: Liu, S., Ghafar, A. B. A., & Herman, S. S. (2026). Urban morphology and the urban heat island effect: a bibliometric review and knowledge mapping of research evolution and thematic trends (2005–2024). Theoretical and Applied Climatology, 157(10), Article 616. https://doi.org/10.1007/s00704-026-06569-1

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06569-1

Keywords: urban heat island, urban morphology, bibliometrics, CiteSpace, land surface temperature, sky view factor, street canyon, thermal comfort, Local Climate Zones, urban planning, remote sensing, research collaboration

The study employed two of the most widely used bibliometric mapping tools in science: CiteSpace and VOSviewer. These software platforms allow researchers to transform vast bibliographic databases into visual networks, revealing patterns that would be invisible to anyone reading papers one at a time. CiteSpace, developed originally by Chaomei Chen, excels at detecting bursts of activity in research fronts, identifying pivotal papers that bridge otherwise disconnected bodies of literature, and tracing how keywords and concepts evolve over time. VOSviewer, created by Nees Jan van Eck and Ludo Waltman at Leiden University, constructs similarity-based maps in which publications, authors, or keywords are clustered according to how frequently they cite one another or co-occur in the same documents. By applying both tools to the medical education literature, the team could triangulate their findings, using one platform’s strengths to compensate for the other’s blind spots and thereby producing a more robust picture of the field’s intellectual architecture.

The timing of the analysis is significant. The researchers deliberately framed their window as the past decade, capturing the period in which digital technologies matured from novelty to necessity. Their results show a steady increase in publications throughout the period, but the most striking feature is the surge from 2022 to 2024, which the authors attribute to two converging forces: rapid technological advancement and the COVID-19 pandemic. When the pandemic forced medical schools worldwide to suspend in-person teaching, clinical rotations, and laboratory sessions almost overnight, digital tools stopped being optional supplements and became the only viable channel for education. That forced adoption, the analysis suggests, permanently altered the trajectory of the field, converting temporary workarounds into lasting infrastructure and triggering a wave of research into how well these tools actually work.

Among the key themes the mapping revealed, artificial intelligence-powered personalization stands out as perhaps the most consequential. Adaptive learning platforms, guided by algorithms that adjust content difficulty and pacing to individual student performance, promise to replace the one-size-fits-all lecture model with something closer to a personal tutor for every learner. The bibliometric data show this theme intersecting with work on big data analytics, in which student interaction data feeds back into the system to refine recommendations continuously. Closely related is the theme of deep learning applied to medical imaging, a field that has transformed both clinical practice and the training of future radiologists. Convolutional neural networks can now detect diseases in retinal photographs, chest radiographs, and pathology slides with performance that in some benchmarks rivals experienced clinicians, and medical schools are increasingly expected to teach students how these systems work, what their limitations are, and how to interpret their outputs responsibly.

Virtual reality and haptic simulation form another major cluster in the research landscape. The analysis identified a rich literature on VR-based surgical training, including simulations of cerebral aneurysm clipping with real-time haptic feedback, endodontic microsurgery rehearsal, and smart haptic gloves that let trainees feel resistance as they practice procedures such as external ventricular drain placement. Systematic reviews cited in the study support the educational validity of virtual haptics in surgical simulation, and meta-analyses of virtual patient simulations in health professions education indicate meaningful learning gains. The pandemic accelerated adoption here as well, with studies documenting how virtual and augmented reality sustained the quality of medical education when cadaver laboratories and operating theaters were off limits. Anatomy education, long dependent on physical dissection, has proven a particularly fertile ground for hybrid models that combine virtual environments with traditional laboratory experiences.

Perhaps no theme has generated more attention, or more controversy, than the arrival of large language models. The bibliometric analysis captured the sudden eruption of research on tools such as ChatGPT following their public release, including studies evaluating their performance on the United States Medical Licensing Examination and the Japanese medical licensing examination. Other cited work compared physician and AI chatbot responses to patient questions posted on public forums, examined ChatGPT-assisted teaching in pediatric clinical skills training, and explored language model-powered simulated patients that give medical students automated feedback on their history-taking technique. These studies suggest that generative AI can serve as an interactive teaching partner, offering on-demand explanation, case-based reasoning practice, and round-the-clock availability that human faculty cannot match. Yet the same literature is candid about the pitfalls: models can hallucinate clinical facts, exhibit biases embedded in their training data, and potentially encourage shortcuts that undermine deep learning if students lean on them uncritically.

The analysis also surfaced the structural inequalities that shadow this technological optimism. The researchers highlight disparities in global research capacity, noting that the literature is dominated by contributions from a limited set of countries and institutions while many regions remain marginal participants in shaping the field. Data privacy concerns, ethical questions surrounding algorithmic decision-making, and resource inequality compound the problem. Advanced VR suites and AI platforms are expensive, and institutions in low-resource settings often cannot afford them, raising the prospect that digital education could widen rather than close existing gaps in medical training quality. The cited literature on simulation in low-resource settings and on blockchain-based platforms for secure data sharing reflects early attempts to address these challenges, but the bibliometric evidence suggests such work remains a small fraction of the overall output.

What makes the study’s conclusions notable is their balance. The authors conclude that intelligent digital platforms have been genuinely transformative, particularly in clinical training, adaptive learning, and medical diagnostics simulation, and that these technologies have the potential to revolutionize medical education. At the same time, they insist that realizing that potential equitably requires confronting ethical, technical, and resource-based challenges head-on. Their recommendations for future research are concrete: foster international collaboration so that expertise and infrastructure flow across borders, develop standardized frameworks for evaluating and integrating digital tools into curricula, and create inclusive, low-cost digital tools that can democratize access to high-quality medical education rather than concentrating it among wealthy institutions.

The broader significance of the work lies in its method as much as its findings. Bibliometric mapping of this kind gives educators, funders, and policymakers an evidence-based view of where a fast-moving field has been and where it is heading, replacing anecdote and enthusiasm with measurable trends. As medical schools everywhere grapple with how to train doctors for an era in which AI assists diagnosis, VR simulates surgery, and chatbots answer clinical questions, studies like this one provide the cartography needed to navigate deliberately rather than drift. The decade from 2015 to 2024 turned digital and intelligence education in medicine from a niche interest into a global research enterprise; the coming decade will determine whether its benefits reach every medical student, or only a fortunate few.

Subject of Research: Bibliometric analysis of digital and artificial intelligence technologies in medical education from 2015 to 2024

Article Title: Digital and Intelligence Education in Medicine: A Bibliometric and Visualization Analysis Using CiteSpace and VOSviewer

Article References: Yang, B. X., Zhou, F., Bai, N., Zhou, S., Luo, C., Wang, Q., Wong, A. K. C., & Lin, F. (2025). Digital and Intelligence Education in Medicine: A Bibliometric and Visualization Analysis Using CiteSpace and VOSviewer. Frontiers of Digital Education, 2(1), Article 10. https://doi.org/10.1007/s44366-025-0046-y

Image Credits: AI Generated

DOI: 10.1007/s44366-025-0046-y

Keywords: medical education, artificial intelligence, bibliometrics, CiteSpace, VOSviewer, virtual reality, large language models, deep learning, adaptive learning, simulation, COVID-19, digital health

The study, published in the Journal of Cellular and Molecular Medicine, is a bibliometric analysis covering two decades of scholarly output on autoimmune hepatitis, from 1 January 2004 to 31 December 2024. Rather than measuring molecules or treating patients, bibliometrics measures the flow of scientific communication itself: which countries publish, which institutions collaborate, which journals shape the field, which authors are most cited, and, crucially, which research topics are gaining or losing momentum over time. The authors searched the Web of Science Core Collection for English-language articles and reviews, completed screening on 19 April 2025, and then processed the resulting dataset using three widely respected tools: VOSviewer version 1.6.20 for network construction and visualization, CiteSpace version 6.4.R1 for detecting citation bursts and knowledge structures, and the R package bibliometrix for quantitative summaries of productivity and thematic evolution.

The scale of the underlying literature is striking. The analysis captured 6310 records involving 110 countries, 5717 institutions, 28,429 authors, and 1348 journals. Annual publication output rose markedly after 2020, a surge that reflects both growing clinical interest and the broader post-pandemic acceleration of immunology and hepatology research. The United States, China, the United Kingdom, and Japan emerged as the leading contributors to the field, and two institutions stood out as hubs of international collaboration: the Mayo Clinic in the United States and King’s College Hospital in London. In terms of journals, Liver International published the most records, while Hepatology, one of the field’s flagship outlets, ranked as the most frequently co-cited journal, a signal of how often researchers draw on its content when building new studies. Among individual scientists, Albert J. Czaja and Ansgar W. Lohse featured prominently in both authorship and citation analyses, confirming their status as reference points for the field.

The authors are careful to note an important caveat about these rankings: bibliometric prominence reflects research activity and network position, not necessarily methodological quality or clinical impact. A highly cited paper is one that other scientists engage with, whether they are confirming, extending, or challenging it. Productivity measures the volume of output, not the value of each contribution. This distinction matters because the real story of the analysis lies not in the league tables of countries and authors but in how the questions attracting scientific attention have shifted across the twenty-year window.

Earlier research on autoimmune hepatitis, the keyword and citation analyses show, was dominated by a well-established trio of concerns: diagnosis, immunosuppressive treatment, and liver transplantation. Those themes formed the backbone of the field’s first decades, matching the clinical priorities of an era when the main challenges were recognizing the disease, calming the immune assault with steroids and azathioprine, and transplanting patients whose livers had already failed. Recent burst and trend analyses, however, tell a different story. The fastest-growing and most attention-grabbing topics now include the gut microbiota, the gut–liver axis, biomarker discovery, immune regulation, multi-omics technologies, and computational approaches such as machine learning. In other words, the field is migrating from descriptive and treatment-oriented questions toward mechanistic and data-intensive ones.

The microbiome pivot is grounded in a growing body of human and animal evidence. In a clinical cohort study, Liwinski and colleagues identified reduced microbial diversity in patients with autoimmune hepatitis and a disease-specific decline in Bifidobacterium, a genus of bacteria often associated with gut health. Notably, lower abundance of these organisms was associated with failure to achieve remission, hinting that gut ecology may track with, and possibly influence, treatment response. Building on that theme, Zhang and colleagues reported increased intestinal permeability among patients, a leakiness of the gut barrier that could allow microbial products to cross into the portal circulation and provoke hepatic inflammation. In murine models, the same team showed that barrier dysfunction and bacterial translocation intensified liver inflammation linked to RIP3, a signaling protein involved in a form of inflammatory cell death, while antibiotic treatment attenuated the liver injury. Together, these studies sketch a plausible pathway from gut barrier failure to immune-mediated liver damage.

Yet the authors of the bibliometric analysis are deliberately cautious about how far this narrative can be pushed. Human associations and animal interventions, they stress, do not establish a causal gut mechanism in patients. Association studies cannot determine whether microbial changes drive the disease or merely accompany it, and findings from murine models of altered permeability and bacterial translocation may not translate cleanly to the complexity of human autoimmune hepatitis. What the current evidence does support, they argue, is further systematic investigation of host–microbe interactions, ideally through designs capable of testing causality rather than simply cataloging correlations.

A similarly tempered picture emerges for biomarker research, one of the field’s hottest emerging areas. A recent review catalogued candidate gene-expression, protein, metabolite, and immune-cell markers but emphasized that clinically useful predictors remain inadequately established. Primary studies have produced intriguing candidate signals: serum metabolite signatures associated with cirrhosis, whole-blood transcriptional differences and candidate fibrosis-linked genes, and differential protein abundance measured in a pediatric cohort. Each of these findings represents a lead worth pursuing, the analysis concludes, but none yet constitutes a validated clinical test that a physician could order to predict disease course or treatment response. The gap between discovery and clinical utility remains wide, and closing it will require independent replication in large, well-characterized cohorts.

The same caution applies to the wave of artificial intelligence and machine learning now reaching autoimmune hepatitis. One deep-learning analysis included 123 pretreatment liver biopsies, while an exploratory machine-learning model was built with 233 development patients and validated in a cohort of only 33 patients. These are small numbers by the standards of clinical prediction modeling, and the studies were conducted in single-center, retrospective settings. Limited cohort sizes and constrained designs reduce the transportability of such models to the diverse patient populations seen in real-world practice, where biopsy availability, staining protocols, and clinical data structures vary widely. The bibliometric authors frame these computational efforts as preliminary evidence, promising but unproven, rather than as tools ready for the clinic.

The study closes with a clear-eyed statement of what bibliometrics can and cannot do, and with a roadmap for the field. Publication counts, network centrality, and co-citation patterns measure scholarly communication; they cannot establish biological causality, diagnostic performance, or treatment effectiveness, and they should never serve as proxies for study quality or patient benefit. What the analysis does reveal is directional change: autoimmune hepatitis research is moving toward microbiome science, biomarker discovery, and computational modeling, but these priorities remain exploratory and demand coordinated validation. The authors call for multicenter prospective studies that integrate standardized clinical phenotypes, treatment exposure, histology, and outcomes with immunomic, microbiome, and multi-omics data, with sampling and analyses prespecified, candidate mechanisms and biomarkers validated independently, and experimental testing where appropriate. Shared standards, they argue, are the only reliable way to distinguish reproducible biology from transient research hotspots and to translate robust signals into tools that genuinely improve diagnosis and care for patients living with this stubborn immune-mediated liver disease.

Subject of Research: Bibliometric analysis of global autoimmune hepatitis research trends from 2004 to 2024

Article Title: Evolving Trends in Autoimmune Hepatitis Research: A Bibliometric Analysis From 2004 to 2024

Article References: Kong, Y., & Lin, Z. (2026). Evolving Trends in Autoimmune Hepatitis Research: A Bibliometric Analysis From 2004 to 2024. Journal of Cellular and Molecular Medicine, 30(17), Article e71348. https://doi.org/10.1111/jcmm.71348

Image Credits: AI Generated

DOI: 10.1111/jcmm.71348

Keywords: autoimmune hepatitis, bibliometrics, gut microbiota, gut–liver axis, biomarkers, multi-omics, machine learning, immunosuppression, Web of Science, VOSviewer, CiteSpace, hepatology