Dual-Database Bibliometric Analysis Combined with Gephi-Based Network Visualization of Artificial Intelligence Applications in the Identification and Diagnosis of Thyroid Space-Occupying Lesions.
Authors
Affiliations (2)
Affiliations (2)
- Clinical Anatomy and Reproductive Medicine Application Institute, The First Affiliated Hospital, Hengyang Medical School, University of South China, Hengyang 421001, China.
- Affiliated Hengyang Hospital of Hunan Normal University & Hengyang Central Hospital, Hengyang, 421001, Hunan, China.
Abstract
Accurate differentiation between benign and malignant thyroid space-occupying lesions is critical for clinical decision-making and early treatment planning. Existing diagnostic methods, including ultrasound and fine-needle aspiration biopsy, are constrained by observer dependence and procedural invasiveness. Recent advances in Artificial Intelligence (AI) have provided objective and non-invasive alternatives for thyroid lesion assessment. This study aimed to delineate the research landscape, identify major hotspots, and summarize recent progress in the application of AI to thyroid space-occupying lesions. Publications issued between 2006 and 2025 were retrieved from WOSCC and PubMed, yielding 1546 records for screening. Bibliometric analyses were conducted using CiteSpace and VOSviewer, and network visualization was performed using Gephi. Country and institutional contributions, author collaborations, journal distribution, keyword co-occurrence, and knowledge unit co-occurrence networks were systematically examined. Publication output increased sharply, with 86% of the included studies published within the past 5 years. China (371 articles, 56.9%) and the United States (109 articles, 16.7%) were the leading contributors and central nodes in the collaboration networks. Shanghai Jiao Tong University, the Chinese Academy of Sciences, and Sun Yat-sen University were the main contributing institutions, and their research efforts mainly involved clinical problems, algorithm development, and global multicenter validation, respectively. Major hotspots included deep learning-based thyroid nodule detection and classification, multimodal feature fusion, model optimization, and clinical risk stratification. Keyword analysis identified deep learning, classification, and risk stratification as core themes. AI research in thyroid diagnosis is progressing from technical exploration to clinically oriented applications. China led the publication output, whereas the United States showed stronger international collaboration. The field has also shifted from simple classification to integrated risk stratification and multimodal analysis, reflecting the increasing alignment with precision medicine. This study outlines the bibliometric profile of AI applications in thyroid space-occupying lesion identification and identifies gaps in collaboration and topic diversity. Broader cross-national cooperation may enhance research quality and support future research and clinical practice.