Keyword Usage Research Articles

Comparative bibliometric analysis of artificial intelligence-assisted polyp diagnosis and AI-assisted digestive endoscopy: trends and growth in AI gastroenterology (2003-2023).

Artificial intelligence is already widely utilized in gastroenterology. This study aims to comprehensively evaluate the research hotspots and development trends within the field of AI in gastroenterology by employing bibliometric techniques to scrutinize geographical distribution, authorship, affiliated institutions, keyword usage, references, and other pertinent data contained within relevant publications. This investigation compiled all pertinent publications related to artificial intelligence in the context of gastrointestinal polyps and digestive endoscopy from 2003 to 2023 within the Web of Science Core Collection database. Furthermore, the study harnessed the tools CiteSpace, VOSviewer, GraphPad Prism and Scimago Graphica for visual data analysis. The study retrieved a total of 2,394 documents in the field of AI in digestive endoscopy and 628 documents specifically related to AI in digestive tract polyps. The United States and China are the primary contributors to research in both fields. Since 2019, studies on AI for digestive tract polyps have constituted approximately 25% of the total AI digestive endoscopy studies annually. Six of the top 10 most-cited studies in AI digestive endoscopy also rank among the top 10 most-cited studies in AI for gastrointestinal polyps. Additionally, the number of studies on AI-assisted polyp segmentation is growing the fastest, with significant increases in AI-assisted polyp diagnosis and real-time systems beginning after 2020. The application of AI in gastroenterology has garnered increasing attention. As theoretical advancements in AI for gastroenterology have progressed, real-time diagnosis and detection of gastrointestinal diseases have become feasible in recent years, highlighting the promising potential of AI in this field.

Social Media Attention to Geopolitical Conflicts - An Analysis of Weibo Users' Comments on the Israeli-Palestinian Conflict

In the globalized information age, social media has become a primary channel for news and information dissemination, particularly during geopolitical conflicts. This study investigates public sentiment and discourse on the Israeli-Palestinian conflict on the Chinese social media platform Weibo. Utilizing content analysis, the researcher conducted sentiment statistics and word frequency analysis on Weibo comments to understand Chinese public attitudes toward this conflict. The research reveals a significant increase in negative emotions from 33.33% in October 2023 to 100% in April 2024, indicating growing public discontent and concern as the conflict intensified. Concurrently, positive emotions sharply declined from 47.62% to 0%, reflecting diminished hopes for a peaceful resolution. Neutral sentiments also fluctuated, initially at 19.05%, dropping to 0% by April 2024. Additionally, the study identifies a shift in keyword usage from "world peace" and "hope" to specific entities like "Israel" and "Hamas," and terms like "disaster." This highlights a change in public and media focus from peace initiatives to the humanitarian impact of the conflict. Understanding these dynamics is crucial for promoting a more inclusive gaming environment, challenging existing gender stereotypes, and fostering social stability. This research contributes to a deeper understanding of Chinese public opinion on international geopolitical issues and underscores the importance of social media in shaping public discourse.

Biomass and organic waste conversion for sustainable bioenergy: A comprehensive bibliometric analysis of current research trends and future directions

The rising demand for renewable energy sources has fueled interest in converting biomass and organic waste into sustainable bioenergy. This study employs a bibliometric analysis (2013-2023) of publications to assess trends, advancements, and future prospects in this field. The analysis explores seven key research indicators, including publication trends, leading contributors, keyword analysis, and highly cited papers. We begin with a comprehensive overview of biomass as a renewable energy source and various waste-to-energy technologies. Employing Scopus and Web of Science databases alongside Biblioshiny and VOSviewer for analysis, the study investigates publication patterns, citation networks, and keyword usage. This systematic approach unveils significant trends in research focus and identifies prominent research actors (countries and institutions). Our findings reveal a significant increase in yearly publications, reflecting the growing global focus on biomass and organic waste conversion. Leading contributors include China, the United States, India, and Germany. Analysis of keywords identifies commonly used terms like "biofuels," "pyrolysis," and "lignocellulosic biomass." The study concludes by proposing future research directions, emphasizing advanced conversion technologies, integration of renewable energy sources, and innovative modelling techniques.

Evaluating the Websites of Academic Departments Using the GRA Method

The academic world heavily relies on search engine referrals, making SEO particularly relevant in this sector. To effectively improve page rankings, SEO encompasses two primary categories of factors: 'on-page' and 'off-page. ‘The on-page' factors are elements directly controlled on the website itself. These include strategic keyword usage, content quality, meta tags, URL structure, and internal linking. Conversely, 'off-page' factors involve external elements that influence a website's ranking. Examples include acquiring quality backlinks, maintaining a strong social media presence, and managing online reputation. To attain the maximum benefits from SEO, it is essential to consider relevant factors and criteria. Employing MCDM techniques allows website owners to evaluate and prioritize various SEO elements effectively, enabling a strategic and data-driven approach to improve their web content's search engine rankings. Today, with the advancement and widespread adoption of information systems, the quantity of websites has risen significantly. According to World Wide Web estimates based on the page index by search engines like Google and Bing, the total number of web pages has reached an impressive 4.48 billion. However, this sheer volume of websites makes it challenging for visitors to promptly find the information they are looking for. Thankfully, search engines play a crucial role in helping users access the relevant information they seek quickly and efficiently. The purpose of this study is to explore the challenges of multiple attribute decision-making when dealing with intuitionistic fuzzy information. In this scenario, the attribute weights are not entirely known, and the attribute values are represented by intuitionistic fuzzy numbers. To determine the attribute weights, an optimization model is constructed based on the traditional Grey Relational Analysis (GRA) method's fundamental principles. The proposed method involves calculating the Grey Relational degree between each alternative and the positive-ideal solution and negative-ideal solution. This degree is then used to define a relative relational degree, which enables the ranking of all alternatives simultaneously with respect to both the positive-ideal solution (PIS) and negative-ideal solution (NIS). Alternative taken as Performance criteria(C1), Design criteria (C2), Content criteria (C3), Meta tags criteria (C4), Backlink criteria (C5). Evaluation preference taken as Abdullah Gul University, Turkey (A1); Adana Science and Technology University, Turkey (A2); Aksaray University, Turkey (A3); Alanya Alaaddin Keykubat University, Turkey (A4); Anadolu University, Turkey (A5).

Pengaruh Penggunaan Keywords Pada Penamaan Listing Airbnb Terhadap Tingkat Popularitas Di Kota Bangkok

This study aims to explore the impact of keyword usage in Airbnb listing names on their popularity in Bangkok. Using regular expression (re) and tokenization methods, we identified the top 100 keywords from the listing name column. These keywords were then categorized based on business knowledge. Subsequently, the relationship between keyword usage and popularity was analyzed using the chi-square test, with popularity measured by the number of reviews in the last 12 months. The data used were sourced from Open Data Airbnb and underwent data cleaning and exploratory data analysis (EDA). The results of this study are expected to provide insights for Airbnb hosts to enhance the appeal of their listings through effective naming strategies.

Ecoanxiety and mental health unveiled: a bibliometric analysis

Ecoanxiety, which encompasses the psychological impacts of environmental change, has emerged as a pressing global concern. However, the complex interrelationship between environmental factors and mental health in the context of ecoanxiety remains underexplored. This bibliometric analysis examines the evolution of international research on ecoanxiety and mental health using the Scopus and Web of Science databases. Parameters analyzed include publication trends over time, contributing countries, research foci, and keyword frequencies related to climate change, ecoanxiety, mental health, and solastalgia. The findings reveal surging scholarship in recent years, exponential publication growth, and increasing international collaborations. The total of 214 documents initially retrieved, 122 peer-reviewed publications met the inclusion criteria after pre-processing. The analysis provides a comprehensive overview of the current state of ecoanxiety and mental health research. It elucidates patterns in the emergence and progression of this burgeoning field to inform future research directions. Specifically, the elucidated features regarding keyword usage and research trends establish a foundation to advance investigations on the nexus between environmental issues and psychological well-being. This bibliometric study synthesizes existing knowledge and unveils fruitful avenues to progress understanding of the psychological ramifications of ecological crises.

Research trends in lipid-lowering therapies for coronary heart disease combined with hyperlipidemia: a bibliometric study and visual analysis.

Cardiovascular disease (CVD) poses a significant global health and economic challenge, with atherosclerosis being a primary cause. Over the past 40years, substantial research has been conducted into the prevention and reversal of atherosclerosis, resulting in the development of lipid-lowering agents such as statins and fibrates. Despite the extensive literature and formulation of numerous therapeutic guidelines in this domain, a comprehensive bibliometric analysis of the current research landscape and trends has not been performed. This study aimed to elucidate the evolution and milestones of research into lipid-lowering treatments for coronary heart disease (CHD) in conjunction with hyperlipidemia through bibliometric analysis, offering insights into future directions for treatment strategies. This study examined publications from 1986 to 2023 retrieved from the Web of Science database (Core Collection). Utilizing tools such as VOSviewer, Pajek, and CiteSpace, we analyzed publication and citation numbers, H-indexes, contributions by countries and institutions, authorship, journal sources, and keyword usage to uncover research trajectories and areas of focus. Our analysis of 587 publications revealed a recent surge in research output, particularly post-2003. The American Journal of Cardiology published the highest number of studies, with 40 articles, whereas Circulation received the highest number of citations (6,266). Key contributors included the United States, Japan, and China, with the United States leading in citation numbers and the H-index. Harvard University and Leiden University emerged as pivotal institutions, and Professors J. Wouter Jukema and Robert P. Giugliano were identified as leading experts. Keyword analysis disclosed five thematic clusters, indicating a shift in research towards new drug combinations and strategies, signaling future research directions. The last 4decades have seen a notable rise in publications on lipid-lowering therapies for CHD and hyperlipidemia, with the United States retaining world-leading status. The increase in international collaboration aids the shift towards research into innovative lipid-lowering agents and therapeutic approaches. PCSK9 inhibitors and innovative combination therapies, including antisense oligonucleotides and angiopoietin-like protein 3 inhibitors, provide avenues for future research, intending to maximize the safety and efficacy of treatment approaches.

Critical Science: A systematic literature review of empirical research (1979–2022)

AbstractThe Critical Science Approach (CSA) was developed in 1979 with the intent to create a consistent conceptual identity unifying the Family and Consumer Sciences (FCS) profession's philosophy, knowledge, and practice. This systematic literature review examines empirical CSA research published between 1979 and 2022 to better understand what the CSA means for the profession. A total of 71 articles met the criteria with the CSA highly integrated in 22 of those articles. Results illustrate a need for increased CSA scholarship, continuity in keyword usage within the CSA language, and collaboration among preprofessional programs and professional organizations to support and sustain a CSA paradigm shift.

Hydrogen energy storage integrated grid: A bibliometric analysis for sustainable energy production

Hydrogen energy storage and grid integration are emerging as key technologies for efficient energy generation and decarbonization, addressing the unpredictability of renewable sources like wind and solar. Hydrogen serves as a flexible energy carrier, enabling storage and conversion back to energy, and supporting various applications. Since there is a lack of bibliometric investigation in the grid-connected hydrogen energy storage system, this review conducted the bibliometric analysis and analyzed the role of hydrogen in enhancing grid-connected systems' sustainability and flexibility, based on data from the Scopus database. The principal search terms that have been used to identify the publications are “Hydrogen AND Storage AND ("Grid-connected" OR "On-grid" OR “Grid-tied” OR “Integrated grid” OR “Grid integration”)” and several keywords have also been utilized in the filter including “Hydrogen Storage”, “Hydrogen”, “Hydrogen Production” and “Grid-connected”. The chosen documents type is only comprised of ‘Article’ and ‘Review’. By conducting an analysis of publication trends, citation patterns, co-authorship, and keyword usage, significant and informative observations regarding the development of grid-connected hydrogen storage system research from 2013 to 2023 (up until November) were unveiled. Notably, 86.55% of the articles focused on simulation study, while the rest are focusing in experimental work, reviews, and technical overview of integrated-grid hydrogen storage system. China is establishing itself as prominent participant, while International Journal of Hydrogen Energy is recognized as a notable medium for distributing research on the integrated-grid hydrogen storage system. Compressed hydrogen tank emerges as a prominent type of storage because of its less energy requirement to increase the density of hydrogen that allow more efficient transportation for hydrogen. The review also underlines numerous factors, issues, challenges, and difficulties that next-generation hydrogen energy storage production should overcome with regards to system sustainability.

Artificial Intelligence in Telemedicine: A Global Perspective Visualization Analysis.

Background: The use of artificial intelligence (AI) in telemedicine has been a popular topic in academic research in recent years, resulting in a surge of literature publications. This study provides a scientific overview of AI in telemedicine through bibliometric and visualization analysis. Methods: The Web of Science Core Collection was used as the data source, and the search was conducted on June 1, 2023. A total of 2,860 articles and review studies published in English between 2010 and 2023 were included. This study analyzed general information on the field, trends in publication output, countries/regions, authors, journals, influential articles, keyword usage, and knowledge flows between disciplines. The Bibliometrix R package, VOSviewer, and CiteSpace were used for the analysis. Results: The rate of articles published on AI in telemedicine is increasing by ∼42.1% annually. The United States and China are the top two countries in terms of the number of articles published, accounting for 37.1% of the overall publication volume. In addition to AI and telemedicine, machine learning, digital health, and deep learning are the top three keywords in terms of frequency of occurrence. The keyword time trend graph shows that ChatGPT became one of the important keywords in 2023. The analysis of burst detection suggests that mobile health, based on mobile phones, may be a promising area for future research. Conclusions: This study systematically reviewed the development of AI in telemedicine and identified current research hotspots and future research directions. The results will provide impetus for the innovative development of this field.

Time-varying window-based herding detection in the non-fungible token (NFT) marketplace

PurposeThe non-fungible token (NFT) market had undergone dramatic growth and a sudden decline during 2021–2022. The market experienced a surge in prices in late 2021 and early 2022, with NFTs being sold at inflated prices. Despite this, by April 2022, the market underwent a correction, and the prices of NFTs returned to more reasonable levels. This can be a result of imitating the actions or judgments of a larger group, which is not systematically proven yet. Therefore, this study systematically investigates the applicability of herding behavior in the NFT market.Design/methodology/approachThis research employs cross-sectional absolute deviation (CSAD) of returns and ordinary least squares (OLS) to test herding behavior with moving time windows of 10, 20 and 30 days based on the sales data collected from public interface of OpenSea between July 1, 2021 and June 30, 2022. Additionally, NFT-related keyword usage analysis is done for the detected herding periods.FindingsAs per the results of the data analyzed, herding behavior was evidenced using 10-, 20- and 30-day time windows from July 1, 2021 to June 30, 2022because of media movement. The findings revealed that this behavior was present and aligned with the overall behavior of the market.Originality/valueThis study introduces CSAD to examine herding behavior patterns within the NFT market. Complementing this method, keyword count-based analysis is employed to identify the underlying causes of herding behavior. Through this comprehensive approach, this study not only uncovers the roots of herding behavior but also offers an assessment of the time windows during which it occurs, considering the plausible socioeconomic contexts that influence these trends.

Kajian Bahasa dalam Konteks Influencer Marketing: Pengaruh Pemakaian Bahasa dalam Pemasaran Akun Selebgram @Fadiljaidi

In a continually evolving digital landscape, influencers play a pivotal role in shaping consumer behaviors. This research focuses on analyzing the language strategies employed by influencers in their marketing content. From the use of creative language styles to tailored keyword usage, this study highlights how language serves as a primary tool in capturing attention, building brand identities, and influencing audience mindsets. Apart from considering the language strategies utilized by influencers, this study also delves into the impact of language usage on their followers. This analysis sparks a discussion on how audiences perceive various types of language used in influencer marketing content. The influence of language on the perception of promoted products or services and purchasing decisions takes center stage in this research.

Bibliometric analysis of the 200 most cited papers in the journal Vojnosanitetski pregled

Background/Aim. Vojnosanitetski pregled (VSP) is the official scientific and professional journal of the University of Defence in Belgrade, Serbia. VSP is a peer-reviewed journal that publishes a wide range of scientific and professional articles. The aim of the study was to perform a bibliometric analysis of the 200 most cited articles published in VSP, and to assess the impact, significance, and scientific contribution of the journal. Methods. Using the Web of Science (WoS) Core Collection (WoSCC) database, 2,664 articles published from 2008 to December 31, 2022, were analyzed, focusing on citation counts, author productivity, and collaboration networks. Results. The analysis revealed that the majority of the most cited articles came from Serbian institutions, with the University of Belgrade being the most prolific contributor. The top five most cited authors as well as the most cited article were identified, and a trend of increasing significance of research on bioactive compounds was noticed. The study also observed a shift in key word usage over time, reflecting changes in research trends within the journal. The collaboration network analysis showed a strong clustering among the leading authors, indicating a collaborative culture that contributes to the journal?s influence. Conclusion. Emphasizing the importance of targeted collaborations makes it evident that maintaining and expanding the journal?s impact on the scientific community is the key to improving medical research. Despite limitations such as reliance on a single database, the findings provide valuable insights into the journal?s role in advancing medical research and highlight the importance of targeted collaboration.

Analisis Big Data Hukum Pariwisata Internasional Periode 2017-2023

This study, conducted from 2017 to 2023, analyzes the relationship between international tourism and law using bibliometric methods on a dataset of 237 selected documents. The findings reveal a declining trend in annual dataset growth, with an average document age of approximately 2.93 years. The study delves into author collaboration, document types, and keyword usage. There was an average of 4,042 citations per document, reflecting the scholarly impact in the field. The dataset included 714 plus keywords and 738 author keywords, contributing to content analysis. Among the 570 unique authors, 87 were sole authors, while 99 documents were single-authored, and the average co-authorship was 2.47. Notably, there was no international co-authorship in the dataset. Document types encompassed articles, books, book chapters, conference papers, and more, with peak scientific production observed in 2020. Treemap and thematic map analysis visualize term distribution and research trends.

Trends in Seed Priming Research in the Past 30 Years Based on Bibliometric Analysis.

Seed priming (SP) treatments are widely used in agriculture and restoration to improve seed germination and seedling vigor. Although there exists a considerable amount of scientific literature on SP, it has seldom undergone visual and quantitative analyses. To gain insights into the patterns observed in SP research over the last three decades, we conducted a bibliometric analysis using the Science Citation Index-Expanded (SCI-E) database, aiming to minimize the similarity score in plagiarism detection. This analysis offers a thorough examination of yearly publications, temporal patterns in keyword usage, the top-performing journals, authors, institutions, and countries within the field of SP. Our research findings suggest a steady annual increase of 10.59% in the volume of SP publications, accompanied by a significant upward trajectory in the average citations received per paper annually. According to the analysis of keywords, it was found that "priming" and "germination" emerged as the most frequently used terms in the field of SP research. Seed Science and Technology ranked first among the top journals, and Plant Physiology had greater influence in the field of SP in terms of number of citations. The majority of the top 10 productive institutions were situated in developing countries. In addition, these nations exhibited the highest volume of published works and citations. Our analysis revealed a shift in research focus within the field of SP over the past three decades, transitioning from agricultural science to encompass plant science and environmental science. With the growing recognition of SP's research across different disciplines, there exist abundant prospects for international and interdisciplinary partnerships, collaborative organizations, and progress in this field.

Bibliometric analysis of equity in transportation

This paper aims to map the research landscape in equity and justice in transport and mobility, identifying key publications, authors, institutions, and interdisciplinary connections to inform future research and policy development. A systematic search of scientific articles on equity and justice in transport and mobility was conducted in the Web of Science, resulting in 759 selected articles that were analyzed using bibliometric and network analysis methods. The analysis reveals a growing trend in scientific production since 2013, with the Journal of Transport Geography, Transport Research Part A: Policy and Practice, and the Journal of Transport Policy being the most prominent journals. The most influential authors and institutions are identified, along with the most cited works. Keyword usage trends and thematic analysis show a shift in focus from environmental equity to urban transport, congestion, and accessibility. Cluster analyses highlight focal points and themes, such as accessibility, mobility, social aspects, transport and spatial equity, environmental justice, transport planning, active transport, and methodological evaluations. Co-citation networks reveal three major clusters with differing research focuses: Lukas K. & Martens K.; Litman T. and Currie, G., and Scwanen T. and Cresswell, T. The findings contribute to a better understanding of the existing knowledge in equity and justice in transport and mobility, enabling researchers and policymakers to identify areas requiring further exploration. Additionally, the paper fosters interdisciplinary connections and sets priorities for future research and policy development.

Analyzing the Evolution of Medical Ethics Education: A Bibliometric Analysis of the Top 100 Cited Articles.

Ethics education plays a pivotal role in healthcare by providing professionals and students with the essential competencies to navigate intricate ethical challenges. This study conducts a comprehensive bibliometric analysis of the most-cited articles on ethics education, investigating parameters such as citation count, document types, geographical origin, journal analysis, publication year, author analysis, and keyword usage. The findings reveal a substantial impact characterized by high citation counts and the influence of a prominent publication focusing on the hidden curriculum and structure of medical education. Moreover, the analysis demonstrates a discernible increase in research output since 2000, signaling a growing recognition of the significance of ethics education in the healthcare domain. Notably, specific journals, particularly those dedicated to medical education and ethics, emerge as major contributors in this field, publishing many articles. Renowned authors have made noteworthy contributions, and emerging themes encompass the ethical implications of virtual reality and artificial intelligence in healthcare education. Additionally, undergraduate medical education garners significant attention, emphasizing the importance of establishing ethical values and professionalism early. Overall, this study highlights the imperative of interdisciplinary collaboration and the necessity for effective ethics education programs to equip healthcare professionals with the requisite skills to navigate complex ethical challenges. The findings inform educators, curriculum developers, and policymakers about enhancing ethics education and ensuring the ethical competence of future healthcare practitioners.

The Effectiveness of Search Engine Optimization Practices in Promoting Turkey's Tourism Destinations. A case Study of Tourism Destinations in Istanbul City

This study examined the effectiveness of Search Engine Optimization (SEO) practices in promoting Turkey's tourism destinations, focusing specifically on Istanbul city. Recognizing SEO as a crucial digital marketing tool, the study significantly contributed to the understanding of how SEO influences the visibility and attractiveness of destinations online. Employing a descriptive survey design, the research evaluated SEO practices on a representative sample of tourism-related websites in Istanbul, including official tourism websites, hotel sites, and travel blogs. The websites were analyzed based on established SEO parameters such as keyword usage, meta tags, backlink quality, website speed, and mobile optimization. The study also performed a comprehensive SERP (Search Engine Result Pages) analysis to gauge the visibility of these websites in organic search results. User engagement metrics such as bounce rate, time spent on the site, and conversion rates were studied to understand the effectiveness of the websites in attracting and retaining potential tourists. Furthermore, surveys were conducted with potential tourists to glean insights into their online search behavior, the perceived credibility of high-ranking websites, and their likelihood of visiting a destination based on its online representation. The findings revealed a strong correlation between effective SEO practices, high SERP rankings, and positive user engagement metrics. The study concluded that well-implemented SEO strategies significantly improved the online visibility and credibility of Istanbul's tourism destinations, thereby attracting a greater number of potential tourists. The findings underscored the necessity for local tourism businesses to invest in building SEO expertise and resources to capitalize on the potential of digital marketing in promoting tourism. Keywords: Search engine optimization, turkey's tourism, Istanbul tourism destinations, digital marketing in tourism, online visibility

Decreased community-acquired pneumonia coincided with rising awareness of precautions before governmental containment policy in Japan.

The effectiveness of population-wide compliance to personal precautions (mask-wearing and hand hygiene) in preventing community-acquired pneumonia has been unknown. In Japan, different types of nonpharmaceutical interventions from personal precautions to containment and closure policies (CACPs, e.g. stay-at-home requests) were sequentially introduced from late January to April 2020, allowing for separate analysis of the effects of personal precautions from other more stringent interventions. We quantified the reduction in community-acquired pneumonia hospitalizations and deaths and assessed if it coincided with the timing of increased public awareness of personal precautions before CACPs were implemented. A quasi-experimental interrupted time-series design was applied to non-COVID-19 pneumonia hospitalization and 30-day death data from April 2015 to August 2020 across Japan to identify any trend changes between February and April 2020. We also performed a comparative analysis of pyelonephritis and biliary tract infections to account for possible changes in the baseline medical attendance. These trend changes were then compared with multiple indicators of public awareness and behaviors related to personal precautions, including keyword usage in mass media coverage and sales of masks and hand hygiene products. Hospitalizations and 30-day deaths from non-COVID-19 pneumonia dropped by 24.3% (95% CI 14.8-32.8) and 16.1% (5.5-25.5), respectively, in February 2020, before the implementation of CACPs, whereas pyelonephritis and biliary tract infections did not suggest a detectable change. These changes coincided with increases in indicators related to personal precautions rather than those related to contact behavior changes. Community-acquired pneumonia could be reduced by population-wide compliance to moderate precautionary measures.

Interpretable machine learning text classification for clinical computed tomography reports – a case study of temporal bone fracture

BackgroundMachine learning (ML) has demonstrated success in classifying patients’ diagnostic outcomes in free-text clinical notes. However, due to the machine learning model's complexity, interpreting the mechanism behind classification results remains difficult. MethodsWe investigated interpretable representations of text-based machine learning classification models. We created machine learning models to classify temporal bone fractures based on 164 temporal bone computed tomography (CT) text reports. We adopted the XGBoost, Support Vector Machine, Logistic Regression, and Random Forest algorithms. To interpret models, we used two major methodologies: (1) We calculated the average word frequency score (WFS) for keywords. The word frequency score shows the frequency gap between positively and negatively classified cases. (2) We used Local Interpretable Model-Agnostic Explanations (LIME) to show the word-level contribution to bone fracture classification. ResultsIn temporal bone fracture classification, the random forest model achieved an average F1-score of 0.93. WFS revealed a difference in keyword usage between fracture and non-fracture cases. Additionally, LIME visualized the keywords' contributions to the classification results. The evaluation of LIME-based interpretation achieved the highest interpreting accuracy of 0.97. ConclusionThe interpretable text explainer can improve physicians' understanding of machine learning predictions. By providing simple visualization, our model can increase the trust of computerized models. Our model supports more transparent computerized decision-making in clinical settings.