Background: In the field of traditional Chinese medicine (TCM), diagnostic work based on tongue images to recognize the physical constitution is a process of collecting clinical information, reasoning, and combining the patient's tongue image features with questioning. It is necessary to simulate the recognition of pathological information of tongue images by TCM practitioners and professional dialogue based on tongue image features, which helps to develop an intelligent interactive system for TCM diagnosis.
Objective: This study aimed to develop and validate a vertical model of the TCM domain with TCM's understanding and reasoning capability for tongue images.
Methods: A TongueVLM multimodal large model is designed, which includes a visual encoder module, a modal fusion module, and a language decoder module. First, the visual encoder based on the CLIP-ViT (Contrastive Language-Image Pre-Training With Vision Transformer) pretrained model is used for image patch, dimensionality reduction, and migration learning, which maps the high-dimensional tongue features into low-dimensional language encoding vectors. Further, a modal fusion module with a residual architecture is applied to map visual features to a natural language word embedding space, realizing the conceptual alignment between visual encoding and TCM terminology. Finally, fine-tuning of visual instructions is performed based on the LLaMA (large language model meta artificial intelligence), and a TCM-domain large language model with 7B parameters is trained.
Results: The constructed multimodal dataset has 3 test datasets, and experiments are conducted using 3000 samples from each test dataset, respectively. Experimental results indicate that the TongueVLM model outperforms general-purpose large models on all 3 tasks. On the multimodal test dataset, the TongueVLM model achieved accuracy rates of 79.8%, 78.6%, and 60.7% in evaluation tasks respectively, it achieves 9.1%, 8.4%, and 1.1% in greater accuracy than LLaVA-OneVision, and is 7.5%, 7%, and 5.9% more accurate than Qwen2.5-VL-7B, with the text generation time being around 24 tokens per second.
Conclusions: The TongueVLM model, which achieves tongue image description generation and physical constitution reasoning in TCM, is suitable for the application of a Chinese medicine intelligent diagnosis system.
Background: Multimorbidity has become a major global public health challenge. However, existing research primarily emphasizes the identification of disease patterns at the population level and lacks the capacity to provide predictive insights into individual future pattern membership. Bridging this gap is crucial for personalized prevention and management.
Objective: This study aims to propose an innovative framework that integrates population-level multimorbidity pattern recognition with individual-level predictive modeling, thus advancing multimorbidity research from descriptive analysis to prospective multimorbidity pattern prediction.
Methods: Using longitudinal health follow-up data, we first applied latent transition analysis (LTA) to identify temporally stable multimorbidity patterns. These patterns were subsequently transformed into predictive labels to construct a novel deep learning model, CLA-Net (Cross-Lag Attention Network). CLA-Net is designed to predict individual future multimorbidity patterns by leveraging the complementary strengths of Gated Recurrent Units (GRU) and transformer architectures. It introduces a bitemporal directed cross-attention mechanism to simultaneously capture temporal dependencies and complex feature interactions. We compared CLA-Net against several advanced baselines and conducted ablation studies to validate its architectural components.
Results: In terms of pattern recognition, the LTA identified 5 clinically meaningful multimorbidity patterns: Cardiometabolic-Multisystem, Hypertension-Arthritis, Respiratory-Musculoskeletal, Metabolic Syndrome, and Gastritis-Arthritis. In terms of prediction, experimental results demonstrated that CLA-Net significantly outperformed all baseline models. CLA-Net achieved an accuracy of 0.8352 (SD 0.0048), a precision of 0.8326 (SD 0.0053), a recall of 0.8312 (SD 0.0056), and an F1-score of 0.8319 (SD 0.0051). Notably, it achieved an area under the curve of 0.9293, surpassing baseline models. Ablation studies confirmed the necessity of the dual-branch architecture and the directed cross-attention mechanism, as removing these components resulted in performance declines ranging from 0.93% to 2.50%.
Conclusions: This study extends the scope of LTA beyond descriptive statistical modeling and establishes the scientific value of multimorbidity pattern prediction as an independent research task. By bridging population-level insights with individual-level prediction, the proposed framework provides a data-driven tool for the prospective prediction of future multimorbidity pattern membership conditional on survival, thereby supporting stratified disease management and care planning, rather than general risk stratification for acute or end-stage deterioration. This offers new methodological and practical value for precision medicine and public health policymaking.
Background: Effective secondary use of healthcare data is hindered by fragmentation and a lack of semantic interoperability due to heterogeneous local terminologies. Standardizing clinical terms using SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms) is essential but remains a manual, labor-intensive, and inconsistent process, especially across multiple institutions. Automated, scalable solutions are needed to support reliable mapping and new concept authoring for large-scale research.
Objective: We aimed to develop a large language model (LLM)-assisted tool that streamlines SNOMED CT terminology mapping and concept authoring, which enables seamless, standardized data integration across multi-institutional clinical datasets.
Methods: The mapping pipeline included preprocessing local terms, syntactic and LLM-based vector similarity mapping, and iterative enrichment based on validated results. Translation and semantic representation used GPT-4o (OpenAI). New concepts were authored through a structured postcoordination process, and both the efficiency and quality of authoring (including duplicate rate and Machine Readable Concept Model validation violations) were quantitatively evaluated. Performance was evaluated using diagnostic and surgical procedural terms from 4 major hospital networks (9 university hospitals) in South Korea, with additional usability feedback gathered from clinical terminologists.
Results: Using reference terms, top-5 accuracy for diagnostic mapping reached 98.7%, 89.7%, 98.5%, and 92.8% across the 4 institutions and 99.2%, 82.6%, 98.7%, and 84.7% for surgical procedural mapping. Implementation of the tool reduced manual mapping rates by 30% and overall manual workload by up to 90%. The proposed tool reduced average mapping and new concept creation time by approximately 75%, while decreasing the final mapping table processing time by 90%. New concept authoring errors also decreased, with duplicate concepts reduced by 83% and modeling rule violations by 72%.
Conclusions: This study developed and validated an automated, LLM-assisted SNOMED CT mapping tool that significantly improved efficiency, mapping accuracy, and new concept quality. Limitations include technical integration challenges and dependency on translation quality. Future directions involve leveraging SNOMED CT's ontology structure and knowledge graphs, enhancing sustainability through ongoing maintenance and quality assurance, and further advancing new concept authoring with automated Machine Readable Concept Model rule enforcement and inactivation processes to achieve robust and scalable terminology standardization.
Background: Technology has improved patient care in hospitals, enhancing the overall patient experience. However, digitalization raises questions on effectively integrating technological strategies to ensure assertive communication of information during emergency department (ED) journeys. Keeping patients well-informed boosts their service perception and satisfaction, a factor often neglected by institutions in EDs.
Objective: This paper analyzes relevant studies on technological strategies designed for EDs aimed at improving patient experience, focusing on communication and information access. We analyze the technologies, outcomes, impacts, and challenges of the strategies.
Methods: A scoping review was conducted using the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines and CADIMA tool. Searches were performed in Scopus, PubMed, IEEE Xplore, and CINAHL databases. Articles published from January 2018 to December 2024 were included. Quality appraisal was performed using the Crowe Critical Appraisal Tool version 1.4. Three reviewers independently examined the title and abstract for eligibility based on the inclusion and exclusion criteria.
Results: Sixteen eligible studies were included. Four technological strategy categories were identified: artificial intelligence-based, simulation-based, infrastructure and hardware technologies, and interfaces and information systems. Mobile and web applications were the main technologies adopted in the studies.
Conclusions: Technological strategies hold significant potential to enhance patient experiences in EDs by providing real-time updates on medical status and care progress. However, their effectiveness depends on usability, literacy, and system design. Existing literature highlights the impact and challenges of deploying and using these strategies in EDs. However, no studies have systematically evaluated long-term outcomes or cost-effectiveness across diverse ED settings.
Background: Primary care in Thailand often uses mixed Thai-English free-text documentation for diagnoses and clinical problems, limiting standardization, interoperability, and secondary data use. Clinical terminologies like Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT), a comprehensive reference terminology, can bridge this gap through the use of structured clinical data. Developing and mapping a local user interface terminology (UIT) is one of the key strategies for implementing SNOMED CT in real-world clinical settings.
Objective: This study aimed to develop a Thai UIT derived from frequently used terms in real-world primary care practice, map these terms to SNOMED CT concepts, and evaluate the extent of concept coverage.
Methods: Frequently used clinical terms were extracted from outpatient medical records from the family, emergency, and internal medicine departments using a customized tokenization method, N-gram analysis, and expert review. This process yielded 2054 Thai-specific terms. All terms were normalized and mapped to SNOMED CT through manual expert-driven and semiautomated tools. Unmapped terms were subsequently analyzed to identify mapping barriers and solutions.
Results: Of the 2054 Thai-specific terms, 2012 were successfully mapped to 2041 (97.98%) SNOMED CT concepts, including 1781 (85.50%) fully, 123 (5.90%) broader, 56 (2.69%) narrower, 81 (3.89%) inexact mappings, and 42 (2.02%) remained unmapped. Most mappings were one-to-one (1984), with 28 terms mapped to multiple concepts (57), covering 1486 unique SNOMED CT concepts. The remaining 42 unmapped terms were mostly due to culturally specific expressions or concepts not yet represented in SNOMED CT. These were categorized for potential postcoordination, exclusion, or national extension development.
Conclusions: This study demonstrates the feasibility of developing a Thai UIT mapped to SNOMED CT and describes mapping challenges. The resulting UIT enhances semantic clarity in clinical documentation and supports better interoperability, clinical decision-making, and health data analytics within Thailand's health care system.
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：