BMC Medical Informatics and Decision Making最新文献

Background: Metastasis significantly influences prognosis in thyroid cancer, especially in papillary thyroid carcinoma. With the rise of artificial intelligence (AI) in medical diagnostics, machine learning (ML) and deep learning (DL) models are being increasingly explored for their ability to enhance the early detection of metastatic spread. This systematic review and meta-analysis aimed to evaluate the diagnostic performance of ML and DL algorithms in detecting metastasis in thyroid cancer.

Method: We conducted a comprehensive search of scientific databases, including PubMed, IEEE, Scopus, and Web of Science, covering literature up to July 1st, 2025. This review included studies published in English that used diagnostic models for metastasis in adults with thyroid cancer. Key metrics analyzed were the area under the receiver operating characteristic curve (AUC-ROC) sensitivity, specificity, and the diagnostic odds ratio (DOR) with a 95% confidence interval (CI). Heterogeneity was quantified using I² statistics, and subgroup and moderator analyses were conducted to identify sources of variability. Risk of bias was assessed using the PROBAST tool. Bias risk and concerns were evaluated using the PROBAST checklist. This study was registered with PROSPERO (CRD42024622930).

Results: Thirty-five studies encompassing 162 estimates were included. The pooled sensitivity was 0.747 (95% CI: 0.715-0.775) and specificity was 0.746 (95% CI: 0.706-0.783). The pooled DOR was 9.45 (95% CI: 7.27-12.28), indicating a strong association between AI predictions and actual metastatic status. The overall AUC-ROC was 0.818. Subgroup analysis demonstrated particularly high accuracy in models targeting distant metastasis. ML models showed slightly higher discriminative ability compared to DL models, and robust performance was observed across a variety of cancer subtypes and input data sources. Moderator analysis further confirmed the stability and adaptability of these models under different clinical and technical settings.

Conclusion: ML and DL algorithms demonstrate favorable diagnostic performance in identifying metastasis in thyroid cancer and may serve as supportive tools in clinical decision-making. Their consistent results across different metastasis types and technical settings highlight their potential to complement existing diagnostic approaches. These findings encourage further exploration and refinement of AI-based methods for integration into routine oncologic practice.

Objective: Clustering is widely used to identify meaningful subgroups in biomedical data, but interpretation remains challenging, especially in the absence of ground-truth labels. Moreover, clustering often produces multiple plausible solutions without a single correct answer. Using dementia phenotypes as a case study, we introduce a Formal Explanation Space (FES) to improve interpretability and facilitate comparison across competing cluster solutions.

Methods: We used spectral coclustering and spectral biclustering to cluster a dataset of dementia patients based on clinical phenotypes (signs and symptoms). To enhance interpretability, we constructed an FES to explain algorithm behavior, assess cluster quality, identify influential features, and characterize cluster composition. Although simultaneous clustering is unsupervised, interpretation was aided by diagnostic labels, which we used for external validation of cluster composition.

Results: Spectral coclustering and spectral biclustering each identified five biologically plausible dementia subgroups, though subgroup composition differed by method. The FES provided a structured framework for comparing these divergent outputs.

Conclusions: Clustering complex biomedical data often produces multiple biologically plausible solutions. Retaining and comparing such solutions within a formal explanation space enhances interpretability and supports the discovery of complementary insights across methods.

Background: In Traditional Chinese Medicine Electronic Medical Records (TCM EMRs), symptom descriptions are often semi-structured, and coarse-grained annotation can lead to symptom nesting and information loss. To address these limitations and improve the precision of symptom representation, this study proposes a fine-grained symptom entity annotation system. Its objective is to convert unstandardized symptom expressions into structured data, thereby enhancing the correlation and standardization of symptom information to support intelligent TCM diagnosis and treatment.

Methods: A five-step approach was employed: First, we drafted a fine-grained annotation guideline based on existing research. Second, we annotated symptom entities and iteratively refined the annotations through trial runs, multiple revisions, and evaluations. Third, we trained and assessed Named Entity Recognition (NER) models. Fourth, we extracted relations using predefined rules. Finally, we generated normalized outputs by integrating these rules and manually validated the extraction results.

Results: The study annotated 500 TCM EMRs over five trials, identified 12 entity categories and 10 relation types. The inter-annotator agreement (IAA) F1 scores for entities and relations were 93.56% and 91.23%, respectively. The final corpus comprises 39,097 entities and 41,373 relation pairs, with 15,853 normalized symptom sentences generated through rule-based combination. Compared to prior studies, TCM symptom information utilization (TCM-SIU) increased by 8.24%. The best-performing NER model achieved an F1 score of 92.29%, while rule-based Relation Extraction (RE) attained an F1 score of 88.17%.

Conclusion: The proposed fine-grained symptom annotation system significantly enhances the utilization of symptom information. It effectively mitigates symptom nesting issues, supports comprehensive association, and facilitates structured output, thereby providing robust data for symptom standardization and precise syndrome differentiation.

The integration of clinical research data across various institutions faces hurdles due to differing definitions, inconsistent terminologies, and inadequate support for interoperable metadata. While biomedical ontologies offer valuable tools for structuring clinical data, they have not yet been fully utilized for creating comprehensive metadata descriptors, such as variable semantics, statistical summaries, and governance elements essential for data discovery and alignment. We present the Clinical Metadata Exploration Ontology (CMEO) that builds upon well-established ontologies to provide a cohesive representation of study designs, data elements, exploratory statistics, and data reuse permissions. CMEO facilitates semantic querying for study exploration and comparison of data elements across studies, particularly when individual-level data cannot be shared. We demonstrate its utility using metadata from five studies: four heart-failure studies and one wearable-based type 1 diabetes study. After serializing, we executed SPARQL queries that operationalized study-level discovery, variable alignment across studies, and governance-constrained reuse. This FAIR-compliant, metadata-driven integration across heterogeneous sources enables scalable, privacy-conscious research and underpins federated clinical data exploration.